re PR target/87927 (ICE: segmentation fault with patchable_function_entry attribute...
[official-gcc.git] / gcc / ChangeLog
blob43b854451cc0cf43d7f00d279874bb5a17df32fd
1 2018-11-16  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
3         PR target/87927
4         * target-def.h: Initialize TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP.
5         Add them to the TARGET_ASM_{,UN}ALIGNED_INT_OP structs.
6         * target.def: Enumerate TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP in
7         the byte_op hook.
8         * target.h: Add psi, pdi, pti to struct asm_int_op definition.
9         * targhooks.c (default_print_patchable_function_entry): Assert
10         asm_int_op does not return a NULL string.
11         * varasm.c (integer_asm_op): Return the op for a partial int type
12         when the requested size does not correspond to an integer type.
13         * config/msp430/msp430.c: Initialize TARGET_ASM_{,UN}ALIGNED_PSI_OP.
14         * doc/tm.texi: Regenerate.
16 2018-11-12  Jason Merrill  <jason@redhat.com>
18         * gimplify.c (gimplify_case_label_expr): Handle hot/cold attributes.
20 2018-11-16  Michael Meissner  <meissner@linux.ibm.com>
22         * config/rs6000/constraints.md (wF constraint): Remove power9
23         fusion documentation.  Just document wF's use for power8 fusion.
24         * config/rs6000/predicates.md (p9_fusion_reg_operand): Delete.
25         (fusion_gpr_addis): Delete power9 fusion support.  Change power8
26         fusion support to require the upper 12 bits to be all 0's or all
27         1's.
28         (fusion_gpr_mem_load): Add comment.
29         (fusion_addis_mem_combo_load): Remove power9 fusion support.
30         (fusion_addis_mem_combo_store): Delete.
31         (fusion_offsettable_mem_operand): Delete.
32         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Do not set
33         power8 fusion here.
34         (ISA_3_0_MASKS_SERVER): Delete power9 fusion.
35         (POWERPC_MASKS): Delete power9 fusion.
36         * config/rs6000/rs6000-protos.h (emit_fusion_load_store): Delete.
37         (fusion_p9_p): Delete.
38         (expand_fusion_p9_load): Delete.
39         (expand_fusion_p9_store): Delete.
40         (emit_fusion_p9_load): Delete.
41         (emit_fusion_p9_store): Delete.
42         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Delete power9
43         fusion support.
44         (rs6000_option_override_internal): Set power8 fusion based on
45         whether we are tuning for power8.  Delete power9 fusion support.
46         (rs6000_opt_masks): Delete -mpower9-fusion switch.
47         (emit_fusion_load): Rename emit_fusion_load_store to
48         emit_fusion_load, and drop fusion store support.  Update callers.
49         (emit_fusion_load_store): Likewise.
50         (emit_fusion_gpr_load): Likewise.
51         (fusion_p9_p): Delete.
52         (expand_fusion_p9_load): Delete.
53         (expand_fusion_p9_store): Delete.
54         (emit_fusion_p9_load): Delete.
55         (emit_fusion_p9_store): Delete.
56         * config/rs6000/rs6000.md (UNSPEC_FUSION_P9): Delete.
57         (GPR_FUSION): Delete.
58         (FPR_FUSION): Delete.
59         (power9 fusion peephole2s): Delete.
60         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Delete.
61         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Delete.
62         (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_load): Delete.
63         (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_store): Delete.
64         (fusion_p9_<mode>_constant): Delete.
65         * config/rs6000/rs6000.opt (-mpower9-fusion): Delete undocumented
66         power9 fusion switch.
67         * doc/md.texi (PowerPC constraints): Update wF constraint
68         documentation for power8 fusion only.
70 2018-11-16  Jakub Jelinek  <jakub@redhat.com>
72         PR rtl-optimization/87475
73         * cfgrtl.c (patch_jump_insn): Allow redirection failure for
74         CROSSING_JUMP_P insns.
75         (cfg_layout_redirect_edge_and_branch): Don't ICE if ret is NULL.
77 2018-11-16  Uros Bizjak  <ubizjak@gmail.com>
79         PR target/88051
80         * config/i386/i386.md (floatunsdidf2): Allow only 64bit AVX512F targets.
81         * config/i386/sse.md (UNSPEC_MOVDI_TO_SSE): New UNSPEC.
82         (movdi_to_sse): Rewrite using UNSPEC_MOVDI_TO_SSE unspec.
84 2018-11-16  Jakub Jelinek  <jakub@redhat.com>
86         PR middle-end/88032
87         * optabs.c (expand_binop): For op0_mode use GET_MODE (op0), unless it
88         is VOIDmode, in which case use int_mode.  Similarly for op1_mode.
90 2018-11-16  Eric Botcazou  <ebotcazou@adacore.com>
92         * tree-switch-conversion.h (switch_decision_tree::emit_case_nodes): Add
93         location_t parameter.
94         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
95         (switch_decision_tree::do_jump_if_equal): Likewise.
96         * tree-switch-conversion.c (switch_decision_tree::emit): Pass location
97         of switch statement to emit_case_nodes.
98         (switch_decision_tree::emit_cmp_and_jump_insns): Add LOC parameter and
99         set it on the newly built GIMPLE comparison statement.
100         (switch_decision_tree::do_jump_if_equal): Likewise.
101         (switch_decision_tree::emit_case_nodes): Add LOC parameter and pass it
102         in calls to do_jump_if_equal as well as recursive calls.
104 2018-11-16  Andi Kleen  <ak@linux.intel.com>
106         * config/i386/i386.md: Allow memory operands to ptwrite.
108 2018-11-16  Michael Matz  <matz@suse.de>
110         * system.h (PRsa): New macro.
111         (SIZE_AMOUNT): Cast number to uint64_t.
112         * alloc-pool.h (pool_usage::dump): Don't use %zu but PRsa.
113         (pool_usage::dump_footer): Likewise and also use PRIu64.
114         * bitmap.h (bitmap_usage::dump): Likewise.
115         * ggc-common.c (ggc_usage::dump): Likewise.
116         * ggc-page.c (ggc_print_statistics): Likewise.
117         * input.c (dump_line_table_statistics): Likewise.
118         * mem-stats.h (mem_usage::dump): Likewise.
119         (mem_usage::dump_footer): Likewise.
120         * rtl.c (dump_rtx_statistics): Likewise.
121         * tree-cfg.c (dump_cfg_stats): Likewise.
122         * tree-dfa.c (dump_dfa_stats): Likewise.
123         * tree-phinodes.c (phinodes_print_statistics): Likewise.
124         * tree-ssanames (ssanames_print_statistics): Likewise.
125         * vec.c (vec_usage::dump): Likewise.
126         (vec_usage::dump_footer): Likewise.
128 2018-11-16  Richard Biener  <rguenther@suse.de>
130         PR tree-optimization/88011
131         * tree-vrp.c (extract_range_from_binary_expr): Fix error in
132         replacing set_value_range_to_undefined and
133         set_value_range_to_varying with method calls.
135 2018-11-16  Ilya Leoshkevich  <iii@linux.ibm.com>
137         * config/s390/s390.md
138         (*r<noxa>sbg_<mode>_srl_bitmask): Do not delegate arithmetic to
139         assembler.
140         (*r<noxa>sbg_<mode>_sll): Likewise.
141         (*r<noxa>sbg_<mode>_srl): Likewise.
142         (*r<noxa>sbg_sidi_srl): New pattern.
144 2018-11-16  Jerome Lambourg  <lambourg@adacore.com>
146         * config/vxworks.h (TARGET_VXWORKS_HAVE_CTORS_DTORS): New macro.
147         Default to TARGET_VXWORKS_RTP.
148         (SUPPORTS_INIT_PRIORITY): Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
149         of TARGET_VXWORKS_RTP.
150         * config/vxworksae.h: Also define TARGET_VXWORKS_HAVE_CTORS_DTORS.
151         * config/vxworks.c: Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
152         of TARGET_VXWORKS_RTP to set targetm.have_ctors_dtors.
154 2018-11-15  Sandra Loosemore  <sandra@codesourcery.com>
156         PR c++/25759
157         * doc/extend.texi (Common Type Attributes): Make it explicit
158         that attribute "packed" can apply to C++ classes.
160 2018-11-15  Martin Sebor  <msebor@redhat.com>
162         PR c++/87541
163         PR c++/87542
164         * tree.c (type_argument_type): New function.
165         * tree.h (type_argument_type): Declare it.
166         * gcc/doc/extend.texi (alloc_align): Update and clarify.
167         (alloc_size, nonnull, sentinel): Same.
169 2018-11-15  Andrew Stubbs  <ams@codesourcery.com>
170             Kwok Cheung Yeung  <kcy@codesourcery.com>
172         * tree-vect-stmts.c (vectorizable_store): Don't ICE when
173         int_mode_for_size fails.
174         (vectorizable_load): Likewise.
176 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
178         * doc/ux.texi (Group logically-related diagnostics): Move
179         discussion of auto_diagnostic_group into this new subsection.
180         Give an example of where this grouping is used.
182 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
184         PR other/19165
185         * Makefile.in (OBJS): Move json.o to...
186         (OBJS-libcommon): ...here and add diagnostic-format-json.o.
187         * common.opt (fdiagnostics-format=): New option.
188         (diagnostics_output_format): New enum.
189         * diagnostic-format-json.cc: New file.
190         * diagnostic.c (default_diagnostic_final_cb): New function, taken
191         from start of diagnostic_finish.
192         (diagnostic_initialize): Initialize final_cb to
193         default_diagnostic_final_cb.
194         (diagnostic_finish): Move "being treated as errors" messages to
195         default_diagnostic_final_cb.  Call any final_cb.
196         (default_diagnostic_finalizer): Add diagnostic_t param.
197         (diagnostic_report_diagnostic): Pass "orig_diag_kind" to
198         diagnostic_finalizer callback.
199         * diagnostic.h (enum diagnostics_output_format): New enum.
200         (diagnostic_finalizer_fn): Reimplement, adding diagnostic_t param.
201         (struct diagnostic_context): Add "final_cb".
202         (default_diagnostic_finalizer): Add diagnostic_t param.
203         (diagnostic_output_format_init): New decl.
204         * doc/invoke.texi (-fdiagnostics-format): New option.
205         * dwarf2out.c (gen_producer_string): Ignore
206         OPT_fdiagnostics_format_.
207         * gcc.c (driver_handle_option): Handle OPT_fdiagnostics_format_.
208         * lto-wrapper.c (append_diag_options): Ignore it.
209         * opts.c (common_handle_option): Handle it.
211 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
213         PR tree-optimization/88015
214         * graphite-isl-ast-to-gimple.c
215         (translate_isl_ast_to_gimple::scop_to_isl_ast): Add missing check
216         for dump_enabled_p.
217         * graphite-sese-to-poly.c (build_poly_scop): Likewise.
219 2018-11-15  Richard Biener  <rguenther@suse.de>
221         PR middle-end/88029
222         * gimple.c (gimple_call_flags): Union flags from decl, type
223         and call fntype.
224         * trans-mem.c (is_tm_pure_call): Simplify.
226 2018-11-15  Richard Biener  <rguenther@suse.de>
228         PR tree-optimization/88031
229         * tree-vect-loop.c (vectorizable_reduction): Move check
230         for multiple types earlier so we get the expected dump.
231         Simplify calls to vectorizable_condition.
232         * tree-vect-stmts.h (vectorizable_condition): Update prototype.
233         * tree-vect-stmts.c (vectorizable_condition): Instead of
234         reduc_def and reduc_index take just a flag.  Simplify
235         code-generation now that we can rely on the defs being set up.
236         (vectorizable_comparison): Remove unused argument.
238 2018-11-15  Richard Biener  <rguenther@suse.de>
240         PR tree-optimization/88030
241         * tree-complex.c (need_eh_cleanup): New global.
242         (update_complex_assignment): Mark blocks that need EH update.
243         (expand_complex_comparison): Likewise.
244         (tree_lower_complex): Allocate and deallocate need_eh_cleanup,
245         perform EH cleanup and schedule CFG cleanup if that did anything.
247 2018-11-15  Jakub Jelinek  <jakub@redhat.com>
249         PR rtl-optimization/88018
250         * cfgrtl.c (fixup_abnormal_edges): Guard moving insns to fallthru edge
251         on the presence of fallthru edge, rather than if it is a USE or not.
253 2018-11-15  Richard Biener  <rguenther@suse.de>
255         PR middle-end/87917
256         * tree-data-ref.c (analyze_miv_subscript): Guard calls to
257         analyze_subscript_affine_affine properly.
259 2018-11-15  Xianmiao Qu  <xianmiao_qu@c-sky.com>
261         * config/csky/csky-linux-elf.h (LINUX_DYNAMIC_LINKER): Remove.
262         (GLIBC_DYNAMIC_LINKER): Define.
263         (LINUX_TARGET_LINK_SPEC): Update the dynamic linker's name.
265 2018-11-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
267         PR tree-optimization/84648
268         * tree-ssa-loop-niter.c (adjust_cond_for_loop_until_wrap): New.
269         (number_of_iterations_cond): Adjust exit cond for loop-until-wrap case
270         by calling adjust_cond_for_loop_until_wrap.
272 2018-11-15  Sandra Loosemore  <sandra@codesourcery.com>
274         PR other/56334
275         * doc/extend.texi (Common Function Attributes): Clarify linker
276         restrictions on "aligned" attribute.
277         (Common Variable Attributes): Likewise.  Mention that linker
278         restrictions don't apply to stack-allocated variables.
280 2018-11-15  Joern Rennecke  <joern.rennecke@riscy-ip.com>
282         * gensupport.c (add_predicate_code): Properly handle ZERO_EXTRACT
283         as an lvalue.
285 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
287         PR lto/55102
288         PR lto/56700
289         * doc/invoke.texi (Optimize Options): Remove bad example about
290         interaction between -flto and -O.  Replace it with a note that
291         you need to compile with -O and not just link.  Copy-edit -flto
292         discussion to reduce verbiage and improve flow.
294 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
296         PR middle-end/59658
297         * doc/invoke.texi (Optimize Options): Clarify that -O0 and -Og
298         also suppress many optimizations.  Alphabetize option lists for
299         -O1, -O2, and -Os.  Add list of options disabled with -Og, and
300         correct documentation for those options to say that.
301         * opts.c (default_options_table): Sort table by level and option
302         name, to make it easier to correlate to the manual.
304 2018-11-14  Uros Bizjak  <ubizjak@gmail.com>
306         * config/i386/i386.c (ix86_print_operand_address_as): Simplify
307         printing of the names of segment registers.
309 2018-11-14  Nathan Sidwell  <nathan@acm.org>
311         PR debug/88006
312         PR debug/87462
313         * dwarf2out.c (dwarf2out_finish): Apply resolve_addr to comdat
314         type list.
316 2018-11-14  David Malcolm  <dmalcolm@redhat.com>
318         * Makefile.in (CFLAGS-optinfo-emit-json.o): Add $(ZLIBINC).
320 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
322         P1236R1 - Signed integers are two's complement
323         * doc/invoke.texi (Wshift-overflow): Adjust documentation for c++2a.
325         PR bootstrap/86739
326         * hash-map.h (hash_map::iterator::reference_pair): New class.
327         (hash_map::iterator::operator*): Return it rather than std::pair.
329 2018-11-14  Jeff Law  <law@redhat.com>
331         * optabs.c (expand_binop): Pass INT_MODE to operand_subword_force
332         iff the operand is a constant.
334 2018-11-14  Aldy Hernandez  <aldyh@redhat.com>
336         * gimple-ssa-evrp-analyze.c
337         (evrp_range_analyzer::record_ranges_from_incoming_edge): Rename
338         ignore_equivs_equal_p to equal_p.
339         * ipa-cp.c (meet_with_1): Use equal_p instead of
340         ignore_equivs_equal_p.
341         * ipa-prop.c (ipa_vr_ggc_hash_traits::equal): Same.
342         * tree-vrp.c (value_range::ignore_equivs_equal_p): Remove.
343         (value_range::operator==): Remove.
344         (value_range::operator!=): Remove.
345         (vrp_prop::visit_stmt): Use equal_p.
346         * tree-vrp.h (value_range): Remove operator==, operator!=,
347         ignore_equivs_equal_p.
348         * vr-values.c (update_value_range): Use equal_p.
350 2018-11-14  Michael Matz  <matz@suse.de>
352         PR middle-end/86575
353         * gimplify.c (collect_fallthrough_labels): Add new argument,
354         return location via that, don't modify statements.
355         (warn_implicit_fallthrough_r): Adjust call, don't use
356         statement location directly.
358 2018-11-14  Richard Biener  <rguenther@suse.de>
360         PR middle-end/87985
361         * tree-data-ref.c (split_constant_offset): Add wrapper
362         allocating a cache hash-map.
363         (split_constant_offset_1): Cache results of expanding
364         expressions from SSA def stmts.
366 2018-11-14  Richard Biener  <rguenther@suse.de>
368         PR middle-end/88021
369         * tree-data-ref.c (lambda_matrix_row_add): Change const1 argument
370         to lambda_int.
371         (lambda_vector_mult_const): Likewise.
372         (lambda_matrix_right_hermite): Use lambda_int temporaries.
374 2018-11-14  Wilco Dijkstra  <wdijkstr@arm.com>
375             Jackson Woodruff  <jackson.woodruff@arm.com>
377         PR 71026/tree-optimization
378         * match.pd: Simplify floating point comparisons.
380 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
382         PR rtl-optimization/87817
383         * config/i386/i386.c (ix86_fold_builtin): For _bzhi_u{32,64} if
384         last argument has low 8 bits clear, fold to 0.
386 2018-11-14  Iain Sandoe  <iain@sandoe.co.uk>
388         * tree-vect-data-refs.c (vect_can_force_dr_alignment_p): Cast
389         MAX_OFILE_ALIGNMENT as needed.
391 2018-11-14  Richard Biener  <rguenther@suse.de>
393         PR tree-optimization/88019
394         * graphite-isl-ast-to-gimple.c (ternary_op_to_tree): Rewrite
395         COND_EXPR arguments to non-trapping overflow.
397 2018-11-14  Claudiu Zissulescu  <claziss@synopsys.com>
399         * config/arc/arc.h (reg_class): Reorder registers classes, remove
400         unused register classes.
401         (REG_CLASS_NAMES): Likewise.
402         (REG_CLASS_CONTENTS): Likewise.
403         (FIXED_REGISTERS): Make lp_count fixed.
404         (BASE_REG_CLASS): Remove ACC16_BASE_REGS reference.
405         (PROGRAM_COUNTER_REGNO): Remove.
406         * config/arc/arc.c (arc_conditional_register_usage): Remove unused
407         register classes, use constants for register numbers, remove
408         reg_class_contents references.
409         (arc_process_double_reg_moves): Add asserts.
410         (arc_secondary_reload): Remove LPCOUNT_REG reference, use
411         lra_in_progress predicate.
412         (arc_init_reg_tables): Remove unused register classes.
413         (arc_register_move_cost): Likewise.
414         (arc_preferred_reload_class): Likewise.
415         (hwloop_optimize): Update rtx patterns involving lp_count register.
416         (arc_return_address_register): Rename ILINK1, INLINK2 regnums macros.
417         * config/arc/constraints.md ("c"): Choose between GENERAL_REGS and
418         CHEAP_CORE_REGS.  Former one will be used for LRA.
419         ("Rac"): Choose between GENERAL_REGS and ALL_CORE_REGS.  Former
420         one will be used for LRA.
421         ("w"): Choose between GENERAL_REGS and WRITABLE_CORE_REGS.  Former
422         one will be used for LRA.
423         ("W"): Choose between GENERAL_REGS and MPY_WRITABLE_CORE_REGS.
424         Former one will be used for LRA.
425         ("f"): Delete constraint.
426         ("k"): Likewise.
427         ("e"): Likewise.
428         ("l"): Change it from register constraint to constraint.
429         * config/arc/arc.md (movqi_insn): Remove unsed lp_count constraints.
430         (movhi_insn): Likewise.
431         (movsi_insn): Update pattern.
432         (arc_lp): Likewise.
433         (dbnz): Likewise.
434         (stack_tie): Remove 'b' constraint letter.
435         (R4_REG): Define.
436         (R9_REG, R15_REG, R16_REG, R25_REG): Likewise.
437         (R32_REG, R40_REG, R41_REG, R42_REG, R43_REG, R44_REG): Likewise.
438         (R57_REG, R59_REG, PCL_REG): Likewise.
439         (ILINK1_REGNUM): Renamed to ILINK1_REG.
440         (ILINK2_REGNUM): Renamed to ILINK2_REG.
441         (Rgp): Remove.
442         (SP_REGS): Likewise.
443         (Rcw): Remove unused reg classes.
444         * config/arc/predicates.md (dest_reg_operand): Just default on
445         register_operand predicate.
446         (mpy_dest_reg_operand): Likewise.
447         (move_dest_operand): Use macros instead of constants.
448         * config/arc/arc.opt (mlra): Switch to lra as default.
450 2018-11-14  Richard Biener  <rguenther@suse.de>
452         PR tree-optimization/87974
453         * tree-vect-loop.c (vectorizable_reduction): When computing
454         the vectorized reduction PHI vector type ignore constant
455         and external defs.
457 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
459         PR tree-optimization/87977
460         * tree-ssa-math-opts.c (optimize_recip_sqrt): Don't reuse division
461         stmt, build a new one and replace the old one with it.  Formatting fix.
462         Call release_ssa_name (x) if !has_other_use and !delete_div.
463         (pass_cse_reciprocals::execute): Before calling optimize_recip_sqrt
464         verify lhs of stmt is still def.
466 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
468         PR rtl-optimization/87507
469         * lower-subreg.c (operand_for_swap_move_operator): New function.
470         (simple_move): Strip simple operators.
471         (find_pseudo_copy): Likewise.
472         (resolve_operand_for_swap_move_operator): New function.
473         (resolve_simple_move): Strip simple operators and swap operands.
475 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
477         PR tree-optimization/87898
478         * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Formatting fix.
479         (ipa_simd_modify_function_body): Remove debug stmts where the first
480         argument was changed into a non-decl.
482 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
484         PR rtl-optimization/87899
485         * lra-lives.c (start_living): Update white space in comment.
486         (enum point_type): New.
487         (sparseset_contains_pseudos_p): New function.
488         (update_pseudo_point): Likewise.
489         (make_hard_regno_live): Use HARD_REGISTER_NUM_P macro.
490         (make_hard_regno_dead): Likewise.  Remove ignore_reg_for_conflicts
491         handling.  Move early exit after adding conflicts.
492         (mark_pseudo_live): Use HARD_REGISTER_NUM_P macro.  Add early exit
493         if regno is already live.  Remove all handling of program points.
494         (mark_pseudo_dead): Use HARD_REGISTER_NUM_P macro.  Add early exit
495         after adding conflicts.  Remove all handling of program points and
496         ignore_reg_for_conflicts.
497         (mark_regno_live): Use HARD_REGISTER_NUM_P macro.  Remove return value
498         and do not guard call to mark_pseudo_live.
499         (mark_regno_dead): Use HARD_REGISTER_NUM_P macro.  Remove return value
500         and do not guard call to mark_pseudo_dead.
501         (check_pseudos_live_through_calls): Use HARD_REGISTER_NUM_P macro.
502         (process_bb_lives): Use HARD_REGISTER_NUM_P and HARD_REGISTER_P macros.
503         Use new function update_pseudo_point.  Handle register copies by
504         removing the source register from the live set.  Handle INOUT operands.
505         Update to the next program point using the unused_set, dead_set and
506         start_dying sets.
507         (lra_create_live_ranges_1): Use HARD_REGISTER_NUM_P macro.
509 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
511         * builtins.c: Replace "source_location" with "location_t".
512         * diagnostic-show-locus.c: Likewise.
513         * diagnostic.c: Likewise.
514         * dumpfile.c: Likewise.
515         * gcc-rich-location.h: Likewise.
516         * genmatch.c: Likewise.
517         * gimple.h: Likewise.
518         * gimplify.c: Likewise.
519         * input.c: Likewise.
520         * input.h: Likewise.  Eliminate the typedef.
521         * omp-expand.c: Likewise.
522         * selftest.h: Likewise.
523         * substring-locations.h (get_source_location_for_substring):
524         Rename to...
525         (get_location_within_string): ...this.
526         * tree-cfg.c: Replace "source_location" with "location_t".
527         * tree-cfgcleanup.c: Likewise.
528         * tree-diagnostic.c: Likewise.
529         * tree-into-ssa.c: Likewise.
530         * tree-outof-ssa.c: Likewise.
531         * tree-parloops.c: Likewise.
532         * tree-phinodes.c: Likewise.
533         * tree-phinodes.h: Likewise.
534         * tree-ssa-loop-ivopts.c: Likewise.
535         * tree-ssa-loop-manip.c: Likewise.
536         * tree-ssa-phiopt.c: Likewise.
537         * tree-ssa-phiprop.c: Likewise.
538         * tree-ssa-threadupdate.c: Likewise.
539         * tree-ssa.c: Likewise.
540         * tree-ssa.h: Likewise.
541         * tree-vect-loop-manip.c: Likewise.
543 2018-11-13  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
545         * doc/extend.texi: Fix typo in the weakref description.
547 2018-11-13  Richard Biener  <rguenther@suse.de>
549         PR tree-optimization/86991
550         * tree-vect-loop.c (vect_is_slp_reduction): Delay reduction
551         group building until we have successfully detected the SLP
552         reduction.
553         (vect_is_simple_reduction): Remove fixup code here.
555 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
557         * dumpfile.c (VERIFY_DUMP_ENABLED_P): New macro.
558         (dump_gimple_stmt): Use it.
559         (dump_gimple_stmt_loc): Likewise.
560         (dump_gimple_expr): Likewise.
561         (dump_gimple_expr_loc): Likewise.
562         (dump_generic_expr): Likewise.
563         (dump_generic_expr_loc): Likewise.
564         (dump_printf): Likewise.
565         (dump_printf_loc): Likewise.
566         (dump_dec): Likewise.
567         (dump_dec): Likewise.
568         (dump_hex): Likewise.
569         (dump_symtab_node): Likewise.
570         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
571         Guard dump call with dump_enabled_p.
572         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
573         * graphite-optimize-isl.c (optimize_isl): Likewise.
574         * graphite.c (graphite_transform_loops): Likewise.
575         * tree-loop-distribution.c (pass_loop_distribution::execute): Likewise.
576         * tree-parloops.c (parallelize_loops): Likewise.
577         * tree-ssa-loop-niter.c (number_of_iterations_exit): Likewise.
578         * tree-vect-data-refs.c (vect_analyze_group_access_1): Likewise.
579         (vect_prune_runtime_alias_test_list): Likewise.
580         * tree-vect-loop.c (vect_update_vf_for_slp): Likewise.
581         (vect_estimate_min_profitable_iters): Likewise.
582         * tree-vect-slp.c (vect_record_max_nunits): Likewise.
583         (vect_build_slp_tree_2): Likewise.
584         (vect_supported_load_permutation_p): Likewise.
585         (vect_slp_analyze_operations): Likewise.
586         (vect_slp_analyze_bb_1): Likewise.
587         (vect_slp_bb): Likewise.
588         * tree-vect-stmts.c (vect_analyze_stmt): Likewise.
589         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.
590         (pass_slp_vectorize::execute): Likewise.
591         (increase_alignment): Likewise.
593 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
595         PR ipa/87955
596         * ipa-inline.c (report_inline_failed_reason): Guard calls to
597         cl_target_option_print_diff and cl_optimization_print_diff with
598         if (dump_file).
600 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
602         * doc/invoke.texi (-fsave-optimization-record): Note that the
603         output is compressed.
604         * optinfo-emit-json.cc: Include <zlib.h>.
605         (optrecord_json_writer::write): Compress the output.
607 2018-11-13  Aldy Hernandez  <aldyh@redhat.com>
609         * tree-vrp.c (value_range_base::dump): Dump type.
610         Do not use INF nomenclature for 1-bit types.
611         (dump_value_range): Group all variants to common dumping code.
612         (debug): New overloaded functions for value_ranges.
613         (value_range_base::dump): Remove no argument version.
614         (value_range::dump): Same.
616 2018-11-13  Richard Biener  <rguenther@suse.de>
618         PR tree-optimization/87931
619         * tree-vect-loop.c (vect_is_simple_reduction): Restrict
620         nested cycles we support to latch computations vectorizable_reduction
621         handles.
623 2018-11-13  Martin Liska  <mliska@suse.cz>
625         PR tree-optimization/87885
626         * cfghooks.c (account_profile_record): Rename
627         to ...
628         (profile_record_check_consistency): ... this.
629         Calculate missing num_mismatched_freq_in.
630         (profile_record_account_profile): New function
631         that calculates time and size of a function.
632         * cfghooks.h (struct profile_record): Remove
633         all tuples.
634         (struct cfg_hooks): Remove after_pass flag.
635         (account_profile_record): Rename to ...
636         (profile_record_check_consistency): ... this.
637         (profile_record_account_profile): New.
638         * cfgrtl.c (rtl_account_profile_record): Remove
639         after_pass flag.
640         * passes.c (check_profile_consistency): Do only
641         checking.
642         (account_profile): Calculate size and time of
643         function only.
644         (pass_manager::dump_profile_report): Reformat
645         output.
646         (execute_one_ipa_transform_pass): Call
647         consistency check before clean upand call account_profile
648         after a clean up is done.
649         (execute_one_pass): Call check_profile_consistency and
650         account_profile instead of using after_pass flag..
651         * tree-cfg.c (gimple_account_profile_record): Likewise.
653 2018-11-13  Martin Liska  <mliska@suse.cz>
655         PR sanitizer/87930
656         * config/i386/i386.c (ix86_option_override_internal): Error
657         about usage -mabi=ms and -fsanitize={,kernel-}address.
659 2018-11-13  Andre Vieira  <andre.simoesdiasvieira@arm.com>
661         * config/aarch64/aarch64.c
662         (aarch64_vectorize_preferred_vector_alignment): Change return type to
663         poly_uint64.
664         (aarch64_simd_vector_alignment_reachable): Adapt to preferred vector
665         alignment being a poly int.
666         * doc/tm.texi (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Change
667         return type to poly_uint64.
668         * target.def (default_preferred_vector_alignment): Likewise.
669         * targhooks.c (default_preferred_vector_alignment): Likewise.
670         * targhooks.h (default_preferred_vector_alignment): Likewise.
671         * tree-vect-data-refs.c (vect_calculate_target_alignment): Likewise.
672         (vect_compute_data_ref_alignment): Adapt to vector alignment being a
673         poly int.
674         (vect_update_misalignment_for_peel): Likewise.
675         (vect_enhance_data_refs_alignment): Likewise.
676         (vect_find_same_alignment_drs): Likewise.
677         (vect_duplicate_ssa_name_ptr_info): Likewise.
678         (vect_setup_realignment): Likewise.
679         (vect_can_force_dr_alignment_p): Change alignment parameter type to
680         poly_uint64.
681         * tree-vect-loop-manip.c (get_misalign_in_elems): Learn to construct a
682         mask with a compile time variable vector alignment.
683         (vect_gen_prolog_loop_niters): Adapt to vector alignment being a poly
684         int.
685         (vect_do_peeling): Exit early if vector alignment is not constant.
686         * tree-vect-stmts.c (ensure_base_align): Adapt to vector alignment being
687         a poly int.
688         (vectorizable_store): Likewise.
689         (vectorizable_load): Likweise.
690         * tree-vectorizer.h (struct dr_vec_info): Make target_alignment field a
691         poly_uint64.
692         (vect_known_alignment_in_bytes): Adapt to vector alignment being a
693         poly int.
694         (vect_can_force_dr_alignment_p): Change alignment parameter type to
695         poly_uint64.
697 2018-11-13  Richard Biener  <rguenther@suse.de>
699         PR tree-optimization/87962
700         * tree-vect-loop.c (vect_is_simple_reduction): More reliably
701         detect outer reduction for disqualifying in-loop uses.
703 2018-11-13  Richard Biener  <rguenther@suse.de>
705         PR tree-optimization/87967
706         * tree-vect-loop.c (vect_transform_loop): Also copy PHIs
707         for constants for the scalar loop.
709 2018-11-13  Alan Modra  <amodra@gmail.com>
711         * config/rs6000/linux64.h (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Exclude
712         integer constants when -mcmodel=medium.
714 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
716         * config/arc/arc.h (FUNCTION_PROFILER): Redefine to empty.
717         * config/arc/elf.h (PROFILE_HOOK): Define.
718         * config/arc/linux.h (PROFILE_HOOK): Likewise.
720 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
722         * config/arc/arc.c (hwloop_optimize): Bailout when detecting a
723         jump table data in the text section.
725 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
727         * config/arc/arc.c (arc_eh_return_address_location): Repurpose it
728         to fit the eh_return pattern.
729         * config/arc/arc.md (eh_return): Define.
730         (VUNSPEC_ARC_EH_RETURN): Likewise.
731         * config/arc/arc-protos.h (arc_eh_return_address_location): Match
732         new implementation.
733         * config/arc/arc.h (EH_RETURN_HANDLER_RTX): Remove it.
736 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
738         * common/config/arc/arc-common.c (arc_option_optimization_table):
739         Millicode optimization is default on for size optimizations.
740         * config/arc/arc-protos.h (arc_check_multi): New function.
741         * config/arc/arc.c (RTX_OK_FOR_OFFSET_P): Rearange.
742         (ENTER_LEAVE_START_REG): Define.
743         (ENTER_LEAVE_END_REG): Likewise.
744         (arc_override_options): Disable millicode when long calls option
745         is on.
746         (arc_frame_info): Change it from int to bool.
747         (arc_compute_frame_size): Clean up.
748         (arc_save_restore): Remove.
749         (frame_save_reg): New function.
750         (frame_restore_reg): Likewise.
751         (arc_enter_leave_p): Likewise.
752         (arc_save_callee_saves): Likewise.
753         (arc_restore_callee_saves): Likewise.
754         (arc_save_callee_enter): Likewise.
755         (arc_restore_callee_leave): Likewise.
756         (arc_save_callee_milli): Likewise.
757         (arc_restore_callee_milli): Likewise.
758         (arc_expand_prologue): Reimplement to emit enter/leave
759         instructions.
760         (arc_expand_epilogue): Likewise.
761         (arc_check_multi): New function.
762         * config/arc/arc.md (push_multi_fp): New pattern.
763         (push_multi_fp_blink): Likewise.
764         (pop_multi_fp): Likewise.
765         (pop_multi_fp_blink): Likewise.
766         (pop_multi_fp_ret): Likewise.
767         (pop_multi_fp_blink_ret): Likewise.
768         * config/arc/arc.opt (mmillicode): Update option.
769         (mcode-density-frame): New option.
770         * config/arc/predicates.md (push_multi_operand): New predicate.
771         (pop_multi_operand): Likewise.
772         * doc/invoke.texi (ARC): Update ARC options information.
775 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
777         * config/arc/arc-protos.h (gen_operands_ldd_std): Add.
778         * config/arc/arc.c (operands_ok_ldd_std): New function.
779         (mem_ok_for_ldd_std): Likewise.
780         (gen_operands_ldd_std): Likewise.
781         * config/arc/arc.md: Add peephole2 rules for std/ldd.
783 2018-11-13  Eric Botcazou  <ebotcazou@adacore.com>
785         * toplev.c (output_stack_usage): Turn test on flag_stack_usage into
786         test on stack_usage_file.
787         (lang_dependent_init): Do not open the .su file if generating LTO.
789 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
791         PR rtl-optimization/87918
792         * simplify-rtx.c (simplify_merge_mask): For COMPARISON_P, use
793         simplify_gen_relational rather than simplify_gen_binary.
795 2018-11-13  Richard Biener  <rguenther@suse.de>
797         * tree-ssanames.h (set_range_info): Use value_range_base.
798         (get_range_info): Likewise.
799         * tree-ssanames.c (set_range_info): Likewise.
800         (get_range_info): Likewise.
801         * tree-vrp.c (value_range_base::union_helper): Split
802         out common parts of value_range[_base]::union_.
803         (value_range_base::union_): Update.
804         (value_range::union_): Likewise.
805         (determine_value_range_1): Use value_range_base.
806         (determine_value_range): Likewise.
807         * tree-vrp.h (value_range_base::union_helper): Move ...
808         (value_range::union_helper): ... from here.
810 2018-11-13  Alan Modra  <amodra@gmail.com>
812         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Negate
813         offset for PRE_DEC.
814         (rs6000_secondary_reload_gpr): Don't call find_replacement.
816 2018-11-13  Sandra Loosemore  <sandra@codesourcery.com>
818         PR middle-end/59634
819         * doc/invoke.texi (Optimize Options): Clarify that the
820         l1-cache-line-size, l1-cache-size, and l2-cache-size parameters
821         apply to data cache size.
823 2018-11-13  Alan Modra  <amodra@gmail.com>
825         * config/rs6000/predicates.md (logical_const_operand),
826         (logical_operand): Correct comment.
827         * config/rs6000/rs6000.c (rs6000_emit_prologue): Comment fix.
829 2018-11-13  Alan Modra  <amodra@gmail.com>
831         * config/rs6000/rs6000.md (addsi3_high): Prefix with '*'.
833 2018-11-13  Alan Modra  <amodra@gmail.com>
835         * gcc/doc/md.texi (Machine Constraints): Correct rs6000 h constraint
836         description.
837         * config/rs6000/rs6000.md (movsi_internal1): Delete MT%0 case
838         covered by alternative.
839         (movcc_internal1): Ignore h for register preference.
840         (mov<mode>_hardfloat64): Likewise.
841         (mov<mode>_softfloat): Ignore c, l, h for register preference.
843 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
845         PR preprocessor/47823
846         * doc/cpp.texi (Alternatives to Wrapper #ifndef): Move #pragma once
847         documentation to...
848         (Pragmas): ...here.
849         * doc/extend.texi (Pragmas): Note additional pragmas documented
850         in the CPP manual.
852 2018-11-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
854         PR middle-end/86677
855         PR middle-end/87528
856         * tree-scalar-evolution.c (expression_expensive_p): Make
857         BUILTIN POPCOUNT as expensive when backend does not define it.
859 2018-11-12  Fredrik Noring  <noring@nocrew.org>
861         * config.gcc: Update with-llsc defaults for MIPS r5900.
863 2018-11-12  Martin Liska  <mliska@suse.cz>
865         PR gcov-profile/87442
866         * common.opt: Add -fprofile-filter-files and -fprofile-exclude-files
867         options.
868         * doc/invoke.texi: Document them.
869         * tree-profile.c (parse_profile_filter): New.
870         (parse_profile_file_filtering): Likewise.
871         (release_profile_file_filtering): Likewise.
872         (include_source_file_for_profile): Likewise.
873         (tree_profiling): Filter source files based on the
874         newly added options.
876 2018-11-12  Martin Liska  <mliska@suse.cz>
878         PR target/87903
879         * doc/extend.texi: Add missing values for __builtin_cpu_is and
880         __builtin_cpu_supports for x86 target.
882 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
884         PR middle-end/21110
885         * doc/rtl.texi (Constants): Clarify that mode of "high" doesn't
886         have to be Pmode.
887         (Arithmetic): Likewise for "lo_sum".
889 2018-11-12  Renlin Li  <renlin.li@arm.com>
891         PR target/87815
892         * dse.c (get_stored_val): Add check for compile-time constantness
893         of gap.
895 2018-11-12  Sudakshina Das  <sudi.das@arm.com>
897         * config/arm/arm-cpus.in (armv8_5, sb, predres): New features.
898         (ARMv8_5a): New fgroup.
899         (armv8.5-a): New arch.
900         (armv8-a, armv8.1-a, armv8.2-a, armv8.3-a, armv8.4-a): New options
901         sb and predres.
902         * config/arm/arm-tables.opt: Regenerate.
903         * config/arm/t-aprofile: Add matching rules for -march=armv8.5-a
904         * config/arm/t-arm-elf (all_v8_archs): Add armv8.5-a.
905         * config/arm/t-multilib (v8_5_a_simd_variants): New variable.
906         Add matching rules for -march=armv8.5-a and extensions.
907         * doc/invoke.texi (ARM options): Document -march=armv8.5-a.
908         Add sb and predres to all armv8-a except armv8.5-a.
910 2018-11-12  Richard Biener  <rguenther@suse.de>
912         * tree-vrp.h (value_range[_base]::set): Make public.  Provide
913         overload for single value.
914         (value_range[_base]::set_nonnull): New.
915         (value_range[_base]::set_null): Likewise.
916         (value_range): Document bitmap copying behavior, mark
917         copy constructor and assignment operator deleted.
918         (value_range::move): New.
919         (value_range::set_and_canonicalize): Default bitmap to zero.
920         (set_value_range_to_nonnull): Remove.
921         (set_value_range_to_null): Likewise.
922         (set_value_range): Likewise.
923         (set_value_range_to_value): Likewise.
924         (extract_range_from_unary_expr): Work on value_range_base.
925         (extract_range_from_binary_expr_1): Likewise.  Rename to...
926         (extract_range_from_binary_expr): ... this.
927         * tree-vrp.c (value_range::update): Clear equiv bitmap
928         if required.
929         (value_range::move): New, move equiv bitmap.
930         (value_range_base::set_undefined): Avoid assignment.
931         (value_range::set_undefined): Likewise.
932         (value_range_base::set_varying): Likewise.
933         (value_range::set_varying): Likewise.
934         (set_value_range): Remove.
935         (value_range_base::set): New overload for value.
936         (value_range::set): Likewise.
937         (set_value_range_to_nonnull): Remove.
938         (value_range_base::set_nonnull): New.
939         (value_range::set_nonnull): Likewise.
940         (set_value_range_to_null): Remove.
941         (value_range_base::set_null): New.
942         (value_range::set_null): Likewise.
943         (range_is_null): Work on value_range_base.
944         (range_is_nonnull): Likewise.
945         (ranges_from_anti_range): Likewise.
946         (extract_range_into_wide_ints): Likewise.
947         (extract_range_from_multiplicative_op): Likewise.
948         (extract_range_from_binary_expr): Likewise.  Update for API changes.
949         (extract_range_from_unary_expr): Likewise.  Remove OBJ_TYPE_REF
950         handling.
951         (value_range::intersect_helper): Avoid copy and assignment.
952         (value_range::union_helper): Likewise.
953         (determine_value_range_1): Adjust.
954         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
955         Avoid assignment by using move.
956         (evrp_range_analyzer::record_ranges_from_stmt): Avoid assignment.
957         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
958         Likewise.
959         * tree-ssanames.c (get_range_info): Likewise.
960         * vr-values.h (vr_values::get_vr_for_comparison): Adjust API.
961         * vr-values.c (vr_values::get_value_range): Adjust.
962         (vr_values::update_value_range): Likewise.
963         (symbolic_range_based_on_p): Work on value_range_base.
964         (vr_values::extract_range_from_binary_expr): Use value_range_base.
965         (vr_values::extract_range_from_unary_expr): Likewise.
966         (vr_values::extract_range_from_cond_expr): Avoid assignment.
967         (vr_values::extract_range_from_comparison): Adjust.
968         (vr_values::check_for_binary_op_overflow): Use value_range_base.
969         (vr_values::extract_range_basic): Adjust.
970         (vr_values::adjust_range_with_scev): Likewise.
971         (vr_values::vrp_visit_assignment_or_call): Likewise.
972         (vr_values::get_vr_for_comparison): Change API to avoid
973         assignment and copy construction.
974         (vr_values::compare_name_with_value): Adjust accordingly.
975         (vr_values::compare_names): Likewise.
976         (vr_values::extract_range_from_phi_node): Avoid assignment and
977         bogus in-place modify of equiv bitmap.
978         (vr_values::simplify_bit_ops_using_ranges): Use value_range_base.
979         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Adjust
980         for extract_range_from_unary_expr API change.
981         * ipa-cp.c (ipa_vr_operation_and_type_effects): Likewise.
983 2018-11-12  Eric Botcazou  <ebotcazou@adacore.com>
985         * config/mcore/mcore.h (WORD_REGISTER_OPERATIONS): Remove duplicate.
987 2018-11-12  Richard Biener  <rguenther@suse.de>
989         * tree-vrp.h (value_range_base::symbolic_p,
990         value_range_base::constant_p, value_range_base::zero_p,
991         value_range_base::singleton_p): Move from value_range.
992         (value_range::dump): Add.
993         * gimple-ssa-evrp-analyze.c
994         (evrp_range_analyzer::record_ranges_from_phis): Use set_varying.
995         * ipa-cp.c (ipcp_vr_lattice::print): Use dump_value_range.
996         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
997         Use set_varying.
998         * tree-vrp.c (value_range::symbolic_p): Move to value_range_base.
999         (value_range::constant_p): Likewise.
1000         (value_range::singleton_p): Likewise.
1001         (value_range_base::dump): Add.
1002         (set_value_range_to_undefined): Remove.
1003         (set_value_range_to_varying): Likewise.
1004         (range_int_cst_p): Take value_range_base argument.
1005         (range_int_cst_singleton_p): Likewise.
1006         (value_range_constant_singleton): Likewise.
1007         (vrp_set_zero_nonzero_bits): Likewise.
1008         (extract_range_from_multiplicative_op): Use set_varying.
1009         (extract_range_from_binary_expr_1): Likewise. Use set_undefined.
1010         (extract_range_from_unary_expr): Likewise.
1011         (dump_value_range_base): Change to overload of dump_value_range.
1012         (vrp_prop::vrp_initialize): Use set_varying and set_undefined.
1013         (vrp_prop::visit_stmt): Likewise.
1014         (value_range::intersect_helper): Likewise.
1015         (value_range::union_helper): Likewise.
1016         (determine_value_range_1): Likewise.
1018 2018-11-12  Richard Biener  <rguenther@suse.de>
1020         * tree-vrp.c (set_value_range_to_nonnull): Clear equiv.
1021         (set_value_range_to_null): Likewise.
1022         * vr-values.c (vr_values::extract_range_from_comparison):
1023         Clear equiv for constant singleton ranges.
1025 2018-11-12 Wei Xiao <wei3.xiao@intel.com>
1027         * config/i386/sse.md: Combine VFIXUPIMM* patterns
1028         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1029         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1030         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Remove.
1031         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1032         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1033         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Remove.
1035 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
1037         PR c/69502
1038         * doc/extend.texi (Common Type Attributes): For the align type
1039         attribute, copy language about decreasing alignment from the
1040         corresponding variable attribute.
1042 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
1044         * config/csky/csky.md (*fpuv2_nmulsf3_1, *fpuv2_nmuldf3_1): Handle
1045         -frounding-math.
1047 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
1049         PR c++/43105
1050         * doc/invoke.texi (C++ Dialect Options): Add warning about mixing
1051         -frtti and -fno-rtti code.
1053 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
1055         PR c/26366
1056         * doc/extend.texi (Other Builtins): Document probability associated
1057         with __builtin_expect.
1059 2018-11-11  Uros Bizjak  <ubizjak@gmail.com>
1061         PR target/87928
1062         * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI
1063         instead of (TARGET_64BIT && ix86_abi == MS_ABI).
1064         * config/i386/darwin.h (STACK_BOUNDARY): Ditto.
1065         * config/i386/cygming.h (STACK_BOUNDARY): Remove.
1067 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
1069         * config/csky/csky-linux-elf.h (CC1_SPEC): Support -profile.
1071 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
1073         * config/csky/csky.h (ASM_PREFERRED_EH_DATA_FORMAT): Define.
1075 2018-11-11  Richard Biener  <rguenther@suse.de>
1077         * tree-vrp.h (class value_range_base): New base class for
1078         value_range containing all but the m_equiv member.
1079         (dump_value_range_base): Add.
1080         (range_includes_zero_p): Work on value_range_base.
1081         * tree-vrp.c (value_range_base::set): Split out base handling
1082         from...
1083         (value_range::set): this.
1084         (value_range::set_equiv): New.
1085         (value_range_base::value_range_base): New constructors.
1086         (value_range_base::check): Split out base handling from...
1087         (value_range::check): this.
1088         (value_range::equal_p): Refactor in terms of
1089         ignore_equivs_equal_p which is now member of the base.
1090         (value_range_base::set_undefined): New.
1091         (value_range_base::set_varying): Likewise.
1092         (value_range_base::dump):Split out base handling from...
1093         (value_range::dump): this.
1094         (value_range_base::set_and_canonicalize): Split out base handling
1095         from...
1096         (value_range::set_and_canonicalize): this.
1097         (value_range_base::union_): New.
1098         * ipa-prop.h (struct ipa_jump_func): Use value_range_base *
1099         for m_vr.
1100         * ipa-cp.c (class ipcp_vr_lattice): Use value_range_base
1101         instead of value_range everywhere.
1102         (ipcp_vr_lattice::print): Use dump_value_range_base.
1103         (ipcp_vr_lattice::meet_with): Adjust.
1104         (ipcp_vr_lattice::meet_with_1): Likewise.
1105         (ipa_vr_operation_and_type_effects): Likewise.
1106         (propagate_vr_across_jump_function): Likewise.
1107         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Likewise.
1108         (ipa_get_value_range): Likewise.
1109         (ipa_set_jfunc_vr): Likewise.
1110         (ipa_compute_jump_functions_for_edge): Likewise.
1112 2018-11-10  Sandra Loosemore  <sandra@codesourcery.com>
1114         PR middle-end/65703
1115         * doc/invoke.texi (Optimize Options): Add @opindex entries
1116         for the positive forms of -fno-xxx and -mno-xxx options
1117         that were lacking them.
1119 2018-11-10  Segher Boessenkool  <segher@kernel.crashing.org>
1121         * combine.c (make_more_copies): Only make an intermediate copy if the
1122         dest of a move is a pseudo.
1124 2018-11-09  Maya Rashish  <coypu@sdf.org>
1126         PR target/87221
1127         * config/netbsd-elf.h (NETBSD_STARTFILE_SPEC): Use crtbeginS.o for PIE.
1128         (NETBSD_ENDFILE_SPEC): Use crtendS.o for PIE.
1130 2018-11-09  Sandra Loosemore  <sandra@codesourcery.com>
1132         PR driver/41179
1133         PR middle-end/65703
1134         * doc/invoke.texi (Optimize Options): Clarify default behavior
1135         for -fno-toplevel-reorder, -fno-defer-pop, and -fno-branch-count-reg.
1137 2018-11-09  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1139         PR tree-optimization/87940
1140         * expr.c (string_constant): Don't strip NOPS in subexpressions.
1141         Fold PLUS_EXPR correctly.
1143 2018-11-09  Ilya Leoshkevich  <iii@linux.ibm.com>
1145         PR target/87762
1146         * config/s390/s390.c (s390_safe_relative_long_p): New function.
1147         (annotate_constant_pool_refs): Skip insns which support
1148         relative addressing.
1149         (annotate_constant_pool_refs_1): New helper function.
1150         (find_constant_pool_ref): Skip insns which support relative
1151         addression.
1152         (find_constant_pool_ref_1): New helper function.
1153         (replace_constant_pool_ref): Skip insns which support
1154         relative addressing.
1155         (replace_constant_pool_ref_1): New helper function.
1156         (s390_mainpool_start): Adapt to the new signature.
1157         (s390_mainpool_finish): Likewise.
1158         (s390_chunkify_start): Likewise.
1159         (s390_chunkify_finish): Likewise.
1160         (pass_s390_early_mach::execute): Likewise.
1161         (s390_prologue_plus_offset): Likewise.
1162         (s390_emit_prologue): Likewise.
1163         (s390_emit_epilogue): Likewise.
1165 2018-11-09  Jakub Jelinek  <jakub@redhat.com>
1167         * gimplify.c (gimplify_scan_omp_clauses): Call sorry_at for valid
1168         but unsupported lastprivate with conditional modifier.
1170 2018-11-09  Jeff Law  <law@redhat.com>
1172         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): Handle
1173         unused argument better.  Add gcc_unreachable to silence warning.
1175 2018-11-09  Martin Sebor  <msebor@redhat.com>
1177         PR middle-end/81824
1178         * attribs.c (has_attribute): New helper function.
1179         (decls_mismatched_attributes, maybe_diag_alias_attributes): Same.
1180         * attribs.h (decls_mismatched_attributes): Declare.
1181         * cgraphunit.c (handle_alias_pairs): Call maybe_diag_alias_attributes.
1182         (maybe_diag_incompatible_alias): Use OPT_Wattribute_alias_.
1183         * common.opt (-Wattribute-alias): Take an argument.
1184         (-Wno-attribute-alias): New option.
1185         * doc/extend.texi (Common Function Attributes): Document copy.
1186         (Common Variable Attributes): Same.
1187         * doc/invoke.texi (-Wmissing-attributes): Document enhancement.
1188         (-Wattribute-alias): Document new option argument.
1190 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
1192         * config/arm/parsecpu.awk (/alias/): Tighten invisible alias
1193         matching criteria.  Remove unused array initializer.
1195 2018-11-09  Bill Schmidt  <wschmidt@linux.ibm.com>
1196             Jinsong Ji  <jji@us.ibm.com>
1198         * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix incorrect
1199         constraints by introducing a new temporary.
1200         (_mm_cvtss_si64): Likewise.
1202 2018-11-09  Martin Liska  <mliska@suse.cz>
1204         * common.opt: Add -fipa-stack-alignment flag.
1205         * doc/invoke.texi: Document it.
1206         * final.c (rest_of_clean_state): Guard stack
1207         shrinking with flag.
1209 2018-11-09  Martin Liska  <mliska@suse.cz>
1211         * cgraph.h (ipa_discover_readonly_nonaddressable_vars): Rename
1212         to ...
1213         (ipa_discover_variable_flags): ... this.
1214         * common.opt: Come up with new flag -fipa-reference-addressable.
1215         * doc/invoke.texi: Document it.
1216         * ipa-reference.c (propagate): Call the renamed fn.
1217         * ipa-visibility.c (whole_program_function_and_variable_visibility):
1218         Likewise.
1219         * ipa.c (ipa_discover_readonly_nonaddressable_vars): Renamed to
1220         ...
1221         (ipa_discover_variable_flags): ... this.  Discover
1222         non-addressable variables only with the newly added flag.
1223         * opts.c: Enable the newly added flag with -O1 and higher
1224         optimization level.
1226 2018-11-09  David Malcolm  <dmalcolm@redhat.com>
1228         * json.cc (selftest::test_writing_literals): Fix comment.
1230 2018-11-09  Martin Liska  <mliska@suse.cz>
1232         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not copy
1233         string to a stack buffer.
1234         (aarch64_parse_cpu): Likewise.
1235         (aarch64_parse_tune): Likewise.
1237 2018-11-09  Richard Biener  <rguenther@suse.de>
1239         PR tree-optimization/87953
1240         * tree-vect-loop.c (vectorizable_reduction): For analysis
1241         always pass ops[0] to vectorizable_condition.
1243 2018-11-09  Stafford Horne  <shorne@gmail.com>
1244             Richard Henderson  <rth@twiddle.net>
1245             Joel Sherrill  <joel@rtems.org>
1247         * common/config/or1k/or1k-common.c: New file.
1248         * config/or1k/*: New.
1249         * config.gcc (or1k*-*-*): New.
1250         * configure.ac (or1k*-*-*): New test for openrisc tls.
1251         * configure: Regenerated.
1252         * doc/install.texi: Document OpenRISC triplets.
1253         * doc/invoke.texi: Document OpenRISC arguments.
1254         * doc/md.texi: Document OpenRISC.
1256 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
1258         * config/arm/arm-cpus.in (arm7tdmi): Add an alias for arm7tdmi-s.
1259         (arm7tdmi-s): Delete CPU.
1260         (arm710t): Add aliases for arm720t and arm740t.
1261         (arm720t, arm740t): Delete CPUs.
1262         (arm920t): Add aliases for arm920, arm922t and arm940t.
1263         (arm920, arm922t, arm940t): Delete CPUs.
1264         (arm10tdmi): Add alias for arm1020t.
1265         (arm1020t): Delete CPU.
1266         (arm9e): Add aliases for arm946e-s, arm966e-s and arm968e-s.
1267         (arm946e-s, arm966e-s, arm968e-s): Delete CPUs.
1268         (arm10e): Add aliases for arm1020e and arm1022e.
1269         (arm1020e, arm1022e): Delete CPU.
1270         * config/arm/arm.md (generic_sched): Remove entries that are now
1271         handled by aliases.
1272         (generic_vfp): Likewise.
1273         * config/arm/arm1020e.md: Simplify tuning selection based on alias
1274         changes.
1275         * config/arm/arm-tune.md: Regenerated.
1276         * config/arm/arm-tables.opt: Regenerated.
1278 2018-11-09  Richard Biener  <rguenther@suse.de>
1280         PR tree-optimization/87621
1281         * tree-vect-loop.c (vectorizable_reduction): Handle reduction
1282         op with only phi inputs.
1283         * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h.
1284         (ch_base::copy_headers): Run CSE on copied loop headers.
1285         (pass_ch_vect::process_loop_p): Simplify.
1287 2018-11-09  Alexandre Oliva <oliva@adacore.com>
1289         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust
1290         for non-w64 x86_64 biarch.
1292 2018-11-09  Alexandre Oliva <aoliva@redhat.com>
1294         PR rtl-optimization/86438
1295         * compare-elim.c (try_eliminate_compare): Use SET_SRC instead
1296         of in_b for the compare if in_b is SET_DEST.
1298         PR target/87793
1299         * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject
1300         non-toplevel UNSPEC.
1302 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1304         * tree-vrp.c (value_range::check): Do not access internals
1305         directly.
1306         (value_range::singleton_p): Same.
1307         (value_range::type): Same.
1308         (vrp_finalize): Use value_range API.
1310 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1312         * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly.
1314 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1316         * vr-values.c (vr_values::get_value_range): Use value_range API
1317         instead of piecing together ranges.
1318         (vr_values::update_value_range): Same.
1320 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1322         * gimple-fold.c (size_must_be_zero_p): Use value_range API instead
1323         of performing ad-hoc calculations.
1324         * tree-ssanames.c (set_range_info): New overloaded function
1325         accepting value_range &.
1326         (get_range_info): Same.
1327         * tree-ssanames.h (set_range_info_raw): Remove.
1328         (set_range_info): New prototype.
1329         (get_range_info): Same.
1330         * tree-vrp.h (value_range::null_p): Rename to zero_p.
1331         * tree-vrp.c (value_range::null_p): Same.
1333 2018-11-09  Jan Hubicka  <jh@suse.cz>
1335         * tree.c (fld_type_variant_equal_p): Test user align flag.
1336         (flt_type_variant): Copy user align flag.
1337         (fld_incomplete_type_of): Clear it.
1339 2018-11-09  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
1341         * config/arm/neon.md (div<mode>3): New pattern.
1343 2018-11-08  Andi Kleen  <ak@linux.intel.com>
1345         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New.
1346         (OPTION_MASK_ISA_PTWRITE_UNSET): New.
1347         (ix86_handle_option): Handle OPT_mptwrite.
1348         * config/i386/cpuid.h (bit_PTWRITE): Add.
1349         * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite.
1350         * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64.
1351         * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__.
1352         * config/i386/i386.c (ix86_target_string): Handle ptwrite.
1353         (ix86_option_override_internal): Handle PTA_PTWRITE.
1354         (ix86_valid_target_attribute_inner_p): Define ptwrite.
1355         (def_builtin2): Force UINT64 to be 64bit only.
1356         * config/i386/i386.h (TARGET_PTWRITE): Add.
1357         (TARGET_PTWRITE_P): Add.
1358         (PTA_PTWRITE): Add.
1359         * config/i386/i386.md: Define ptwrite.
1360         * config/i386/i386.opt: Add -mptwrite.
1361         * config/i386/immintrin.h (_ptwrite64): Add.
1362         (_ptwrite32): Add
1363         * doc/extend.texi: Document __builtin_ia32_ptwrite*.
1364         * doc/invoke.texi: Document -mptwrite.
1366 2018-11-08  Peter Bergner  <bergner@linux.ibm.com>
1368         PR rtl-optimization/87600
1369         * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage.
1370         * lra-constraints.c (process_alt_operands): Skip illegal hard
1371         register usage.  Prefer reloading non hard register operands.
1373 2018-11-08  Sandra Loosemore  <sandra@codesourcery.com>
1375         PR other/36572
1376         * doc/invoke.texi (Optimize Options): Clarify default behavior
1377         for -fno-sched-interblock and -fno-sched-spec.
1379 2018-11-08  Roman Geissler  <roman.geissler@amadeus.com>
1381         * collect2.c (linker_select):  Add USE_LLD_LD.
1382         (ld_suffixes): Add ld.lld.
1383         (main): Handle -fuse-ld=lld.
1384         * common.opt (-fuse-ld=lld): New option.
1385         * doc/invoke.texi (-fuse-ld=lld): Document.
1386         * opts.c (common_handle_option): Handle OPT_fuse_ld_lld.
1388 2018-11-08  Paul Koning  <ni1d@arrl.net>
1390         * config/pdp11/constraints.md: Add "Z" series constraints for use
1391         with pre-dec and post-inc addressing.
1392         * config/pdp11/pdp11-protos.m (expand_block_move): Delete.
1393         (pdp11_expand_operands): Add int argument (word count).
1394         (pdp11_sp_frame_offset): Delete.
1395         (pdp11_cmp_length): New function.
1396         (pushpop_regeq): New function.
1397         * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P):
1398         Add hook.
1399         (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new
1400         frame layout.
1401         (pdp11_initial_elimination_offset): Ditto.
1402         (pdp11_expand_operands): Add word count argument.  Bugfixes.
1403         (output_move_multiple): Change how pointer adjustment is done.
1404         (pdp11_gen_int_label): Correct format.
1405         (output_ascii): Ditto.
1406         (pdp11_asm_output_var): Add code for DEC assembler case.
1407         (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer
1408         value.
1409         (legitimate_const_double_p): Ditto.
1410         (pdp11_register_move_cost): Adjust for new register classes.
1411         (pdp11_regno_reg_class): Ditto.
1412         (expand_block_move): Delete.
1413         (pushpop_regeq): New function.
1414         (pdp11_legitimate_address_p): Bugfix in check for constant
1415         offset.
1416         (pdp11_sp_frame_offset): Delete.
1417         (pdp11_reg_save_size): New helper function for new frame layout.
1418         (output_addr_const_pdp11): Remove CONST_DOUBLE case.
1419         (pdp11_expand_shift): Bugfix in check for constant shift count.
1420         (pdp11_shift_length): Ditto.
1421         (pdp11_assemble_shift): Copy input to pdp11_expand_operands.
1422         (pdp11_cmp_length): New function.
1423         * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for
1424         some compile options.
1425         (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM.
1426         (CALL_USED_REGISTERS): Ditto.
1427         (ELIMINABLE_REGS): Ditto.
1428         (REGISTER_NAMES): Ditto.
1429         (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z
1430         constraints.
1431         (REG_CLASS_NAMES): Ditto.
1432         (REG_CLASS_CONTENTS): Ditto.  Also remove
1433         HARD_FRAME_POINTER_REGNUM.
1434         (CPU_REG_CLASS): New macro.
1435         (CLASS_MAX_NREGS): Adjust for new register classes.
1436         (FUNCTION_PROFILER): Make no-op.
1437         (may_call_alloca): Remove unused declaration.
1438         (ASM_OUTPUT_ALIGN): Add workaround for PR87795.
1439         (ASM_OUTPUT_SKIP): Fix format.
1440         * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM.
1441         (HARD_FRAME_POINTER_REGNUM): Remove.
1442         (return): Delete.
1443         (*rts): Rename.  Remove epilogue related checks.
1444         (cmpsi, cmpdi): New insn.
1445         (cbranch<mode>4): Change to apply to SI and DI modes as well.
1446         (mov<mode>): Change constraints to enforce that push/pop
1447         destination cannot use the same register as source.
1448         (*mov<mode><cc_cc>): Ditto.
1449         (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move
1450         at assembly output rather than as RTL expander.
1451         (zero_extendqihi2): Bugfix in check for same registers.
1452         (adddi3_nocc): Bugfix in check for constant operand.
1453         (addsi3_nocc): Ditto.
1454         (subdi3_nocc): Ditto.
1455         (subsi3_nocc): Ditto.
1456         (negdi2_nocc): Copy input to pdp11_expand_operands.
1457         (negsi2_nocc): Ditto.
1458         (bswap2_nocc): Ditto.
1459         * config/pdp11/pdp11.opt (mlra): Fix documentation.
1460         * config/pdp11/t-pdp11: Use -Os.
1462 2018-11-08  Richard Earnshaw  <rearnsha@arm.com>
1464         * config/arm/parsecpu.awk (/alias/): New parsing rule.
1465         (/begin cpu/): Check that the cpu name hasn't been previously defined.
1466         (gen_comm_data): Print out CPU alias tables.
1467         (check_cpu): Match aliases when checking the CPU name.
1468         * config/arm/arm-protos.h (cpu_alias): New structure.
1469         (cpu_option): Add entry for aliases.
1470         * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110
1471         strongarm1100 and strongarm1110.
1472         (strongarm110, strongarm1100, strongarm1110): Delete CPU entries.
1473         (config/arm/arm-generic.md): Remove redundant references to
1474         strongarm110, strongarm1100 and strongarm1110.
1475         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1476         Scan aliases for additional hints.
1477         (arm_parse_cpu_option_name): Also match a cpu name against the list
1478         of aliases.
1479         * config/arm/arm-tables.opt: Regenerated.
1480         * config/arm/arm-tune.md: Regenerated.
1482 2018-11-08  Jakub Jelinek  <jakub@redhat.com>
1484         * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR,
1485         BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT,
1486         BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
1487         BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR,
1488         BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
1489         BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New.
1490         * gengtype.c (open_base_files): Add omp-general.h.
1491         * gimple.c (gimple_build_omp_critical):
1492         (gimple_build_omp_taskgroup): Add CLAUSES argument.  Call
1493         gimple_omp_taskgroup_set_clauses.
1494         (gimple_build_omp_atomic_load): Add mo argument, call
1495         gimple_omp_atomic_set_memory_order.
1496         (gimple_build_omp_atomic_store): Likewise.
1497         (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP.
1498         * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT
1499         instead of GSS_OMP.
1500         (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead
1501         of GSS_OMP_SINGLE_LAYOUT, adjust comments.
1502         * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT
1503         and GF_OMP_ATOMIC_MEMORY_ORDER.  Remove GF_OMP_ATOMIC_SEQ_CST, use
1504         different value for GF_OMP_ATOMIC_NEED_VALUE.
1505         (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to
1506         comments.
1507         (struct gimple_statement_omp_single_layout): And remove here.
1508         (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather
1509         than gimple_statement_omp_single_layout.
1510         (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow
1511         GIMPLE_OMP_TEAMS.
1512         (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise.
1513         (gimple_omp_subcode): Formatting fix.
1514         (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr,
1515         gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg,
1516         gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg,
1517         gimple_omp_teams_host, gimple_omp_teams_set_host,
1518         gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p,
1519         gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr,
1520         gimple_omp_taskgroup_set_clauses): New inline functions.
1521         (gimple_build_omp_atomic_load): Add enum omp_memory_order argument.
1522         (gimple_build_omp_atomic_store): Likewise.
1523         (gimple_omp_atomic_seq_cst_p): Remove.
1524         (gimple_omp_atomic_memory_order): New function.
1525         (gimple_omp_atomic_set_seq_cst): Remove.
1526         (gimple_omp_atomic_set_memory_order): New function.
1527         (gimple_build_omp_taskgroup): Add clauses argument.
1528         * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function.
1529         (dump_gimple_omp_task): Print taskwait with depend clauses.
1530         (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use
1531         dump_omp_atomic_memory_order.
1532         (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP.
1533         * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY,
1534         GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL.
1535         (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags,
1536         renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS,
1537         ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and
1538         ORT_UNTIED_TASKLOOP enumerators.
1539         (enum gimplify_defaultmap_kind): New.
1540         (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and
1541         target_map_pointers_as_0len_arrays members, add defaultmap.
1542         (new_omp_context): Initialize defaultmap member.
1543         (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1544         (maybe_fold_stmt): Don't fold even in host teams regions.
1545         (omp_firstprivatize_variable): Handle ORT_TASKGROUP like
1546         ORT_WORKSHARE.  Test ctx->defaultmap[GDMK_SCALAR] instead of
1547         ctx->omp_firstprivatize_variable.
1548         (omp_add_variable): Don't add private/firstprivate for VLAs in
1549         ORT_TASKGROUP.
1550         (omp_default_clause): Print "taskloop" rather than "task" if
1551         ORT_*TASKLOOP.
1552         (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1553         Handle new defaultmap clause kinds.
1554         (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.  Allow simd
1555         iterator to be lastprivate or private.  Fix up diagnostics if linear
1556         is used on collapse>1 simd iterator.
1557         (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1558         (gimplify_omp_depend): New function.
1559         (gimplify_scan_omp_clauses): Add shared clause on parallel for
1560         combined parallel master taskloop{, simd} if taskloop has
1561         firstprivate, lastprivate or reduction clause.  Handle
1562         OMP_CLAUSE_REDUCTION_TASK diagnostics.  Adjust tests for
1563         ORT_COMBINED_TEAMS.  Gimplify depend clauses with iterators.  Handle
1564         cancel and simd OMP_CLAUSE_IF_MODIFIERs.  Handle
1565         OMP_CLAUSE_NONTEMPORAL.  Handle new defaultmap clause kinds.  Handle
1566         OMP_CLAUSE_{TASK,IN}_REDUCTION.  Diagnose invalid conditional
1567         lastprivate.
1568         (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL.  Handle
1569         GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY.  
1570         (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL.  Handle
1571         OMP_CLAUSE_{TASK,IN}_REDUCTION.
1572         (gimplify_omp_task): Handle taskwait with depend clauses.
1573         (gimplify_omp_for): Add shared clause on parallel for combined
1574         parallel master taskloop{, simd} if taskloop has firstprivate,
1575         lastprivate or reduction clause.  Use ORT_TASKLOOP or
1576         ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK.  Adjust
1577         tests for ORT_COMBINED_TEAMS.  Handle C++ range for loops with
1578         NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS.  Firstprivatize
1579         __for_end and __for_range temporaries on OMP_PARALLEL for
1580         distribute parallel for{, simd}.  Move OMP_CLAUSE_REDUCTION
1581         and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct
1582         sandwiched in between two taskloops.
1583         (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR]
1584         instead of ctx->omp_firstprivatize_variable.
1585         (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or
1586         ORT_COMBINED_HOST_TEAMS if not inside of target construct.  If
1587         host teams, use gimplify_and_return_first etc. for body like
1588         for target or target data constructs, and at the end call
1589         gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object.
1590         (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead
1591         of OMP_ATOMIC_SEQ_CST, pass it as new argument to
1592         gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove
1593         gimple_omp_atomic_set_seq_cst calls.
1594         (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate
1595         case, handle taskgroup clauses.
1596         * lto-streamer-out.c (hash_tree): Handle
1597         OMP_CLAUSE_{TASK,IN}_REDUCTION.
1598         * Makefile.in (GTFILES): Add omp-general.h.
1599         * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND,
1600         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
1601         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
1602         BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START,
1603         BUILT_IN_GOMP_LOOP_DOACROSS_START,
1604         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT,
1605         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
1606         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
1607         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
1608         BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START,
1609         BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START,
1610         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT,
1611         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
1612         BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
1613         BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME,
1614         BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START,
1615         BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER,
1616         BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER,
1617         BUILT_IN_GOMP_TASK_REDUCTION_REMAP,
1618         BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins.
1619         * omp-expand.c (workshare_safe_to_combine_p): Return false for
1620         non-worksharing loops.
1621         (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero.
1622         (determine_parallel_type): Don't combine parallel with worksharing
1623         which has _reductemp_ clause.
1624         (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or
1625         GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime*
1626         if there is nonmonotonic modifier or if there is no modifier and no
1627         ordered clause.  For dynamic and guided schedule without monotonic
1628         and nonmonotonic modifier, default to nonmonotonic.
1629         (expand_omp_for): Likewise.  Adjust expand_omp_for_generic caller, use
1630         GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are
1631         task reductions.
1632         (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if
1633         there are any reduction clauses.
1634         (expand_taskwait_call): New function.
1635         (expand_teams_call): New function.
1636         (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call
1637         expand_teams_call for it.  Formatting fix.  Handle taskwait with
1638         depend clauses.
1639         (expand_omp_for_generic): Add SCHED_ARG argument.  Handle expansion
1640         of worksharing loops with task reductions.
1641         (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle
1642         expansion of worksharing loops with task reductions.
1643         (expand_omp_sections): Handle expansion of sections with task
1644         reductions.
1645         (expand_omp_synch): For host teams call expand_omp_taskreg.
1646         (omp_memory_order_to_memmodel): New function.
1647         (expand_omp_atomic_load, expand_omp_atomic_store,
1648         expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order
1649         instead of gimple_omp_atomic_seq_cst_p.
1650         (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with
1651         depend clauses as a standalone directive.
1652         * omp-general.c (enum omp_requires): New variable.
1653         (omp_extract_for_data): Initialize have_reductemp member.  Allow
1654         NE_EXPR even in OpenMP loops, transform them into LT_EXPR or
1655         GT_EXPR loops depending on incr sign.  Formatting fixes.
1656         * omp-general.h (struct omp_for_data): Add have_reductemp member.
1657         (enum omp_requires): New enum.
1658         (omp_requires_mask): Declare.
1659         * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix.
1660         Fix comment typos.
1661         * omp-low.c (struct omp_context): Add task_reductions and
1662         task_reduction_map fields.
1663         (is_host_teams_ctx): New function.
1664         (is_taskreg_ctx): Return true also if is_host_teams_ctx.
1665         (use_pointer_for_field): Use is_global_var instead of
1666         TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized
1667         in outer contexts.
1668         (build_outer_var_ref): Ignore taskgroup outer contexts.
1669         (delete_omp_context): Release task_reductions and task_reduction_map.
1670         (scan_sharing_clauses): Don't add any fields for reduction clause on
1671         taskloop.  Handle OMP_CLAUSE__REDUCTEMP_.  Handle
1672         OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task
1673         modifier.  Don't ignore shared clauses in is_host_teams_ctx contexts.
1674         Handle OMP_CLAUSE_NONTEMPORAL.
1675         (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if
1676         needed.
1677         (scan_omp_parallel): Add _reductemp_ clause if there are any reduction
1678         clauses with task modifier.
1679         (scan_omp_task): Handle taskwait with depend clauses.
1680         (finish_taskreg_scan): Move field corresponding to _reductemp_ clause
1681         first.  Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present.
1682         Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL.
1683         (scan_omp_for): Fix comment formatting.
1684         (scan_omp_teams): Handle host teams constructs.
1685         (check_omp_nesting_restrictions): Allow teams with no outer
1686         OpenMP context.  Adjust diagnostics for teams strictly nested into
1687         some explicit OpenMP construct other than target.  Allow OpenMP atomics
1688         inside of simd regions.
1689         (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups.
1690         (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump
1691         taskreg_nesting_level while scanning host teams construct.
1692         (task_reduction_read): New function.
1693         (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop
1694         construct.  Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_
1695         clauses.  Handle OMP_CLAUSE_REDUCTION with task modifier.  Remove
1696         second argument create_tmp_var if it is NULL.  Don't ignore shared
1697         clauses in is_host_teams_ctx contexts.  Handle
1698         OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE
1699         clauses.
1700         (lower_reduction_clauses): Ignore reduction clauses with task
1701         modifier.  Remove second argument create_tmp_var if it is NULL.
1702         Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed.
1703         (lower_send_clauses): Ignore reduction clauses with task modifier.
1704         Handle OMP_CLAUSE__REDUCTEMP_.  Don't send anything for
1705         OMP_CLAUSE_REDUCTION on taskloop.  Handle OMP_CLAUSE_IN_REDUCTION.
1706         (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't
1707         rely that it is the last stmt in body so far.  Ignore outer taskgroup
1708         contexts.
1709         (omp_task_reductions_find_first, omp_task_reduction_iterate,
1710         lower_omp_task_reductions): New functions.
1711         (lower_omp_sections): Handle reduction clauses with taskgroup
1712         modifiers.  Adjust maybe_add_implicit_barrier_cancel caller.
1713         (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller.
1714         (lower_omp_for): Likewise.  Handle reduction clauses with taskgroup
1715         modifiers.
1716         (lower_omp_taskgroup): Handle taskgroup reductions.
1717         (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer.
1718         Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses.
1719         (lower_depend_clauses): If there are any
1720         OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET
1721         depend clauses, use a new array format.  If OMP_CLAUSE_DEPEND_LAST is
1722         seen, assume lowering is done already and return early.  Set kind
1723         on artificial depend clause to OMP_CLAUSE_DEPEND_LAST.
1724         (lower_omp_taskreg): Handle reduction clauses with task modifier on
1725         parallel construct.  Handle reduction clause on taskloop construct.
1726         Handle taskwait with depend clauses.
1727         (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams
1728         for host teams constructs.
1729         * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction,
1730         nontemporal and _reductemp_ clause entries.
1731         (omp_clause_code_name): Likewise.
1732         (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION,
1733         OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_.
1734         * tree-core.h (enum omp_clause_code): Add
1735         OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}.
1736         (enum omp_clause_defaultmap_kind, enum omp_memory_order): New.
1737         (struct tree_base): Add omp_atomic_memory_order field into union.
1738         Remove OMP_ATOMIC_SEQ_CST comment.
1739         (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET
1740         and OMP_CLAUSE_DEPEND_DEPOBJ.
1741         (struct tree_omp_clause): Add subcode.defaultmap_kind.
1742         * tree.def (OMP_TASKGROUP): Add another operand, move next to other
1743         OpenMP constructs with body and clauses operands.
1744         * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP.
1745         (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE.
1746         (OMP_TASKGROUP_CLAUSES): Define.
1747         (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of
1748         OMP_CLAUSE__LOOPTEMP_.
1749         (OMP_ATOMIC_SEQ_CST): Remove.
1750         (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE,
1751         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define.
1752         (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT,
1753         OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER,
1754         OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER,
1755         OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle
1756         OMP_CLAUSE_{,IN_,TASK_}REDUCTION.
1757         (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN,
1758         OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY,
1759         OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND):
1760         Define.
1761         * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses.
1762         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
1763         OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL.
1764         (convert_local_omp_clauses): Likewise.  Remove useless test.
1765         * tree-parloops.c (create_call_for_reduction_1): Pass
1766         OMP_MEMORY_ORDER_RELAXED as new argument to
1767         dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store.
1768         * tree-pretty-print.c (dump_omp_iterators): New function.
1769         (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_,
1770         OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION.  Print
1771         reduction modifiers.  Handle OMP_CLAUSE_DEPEND_DEPOBJ and
1772         OMP_CLAUSE_DEPEND_MUTEXINOUTSET.  Print iterators in depend clauses.
1773         Print __internal__ for OMP_CLAUSE_DEPEND_LAST.  Handle cancel and
1774         simd OMP_CLAUSE_IF_MODIFIERs.  Handle new kinds of
1775         OMP_CLAUSE_DEFAULTMAP. Print conditional: for
1776         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL.
1777         (dump_omp_atomic_memory_order): New function.
1778         (dump_generic_node): Use it.  Print taskgroup clauses.  Print
1779         taskwait with depend clauses.
1780         * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare.
1781         * tree-streamer-in.c (unpack_ts_omp_clause_value_fields):
1782         Handle OMP_CLAUSE_{TASK,IN}_REDUCTION.
1783         * tree-streamer-out.c (pack_ts_omp_clause_value_fields,
1784         write_ts_omp_clause_tree_pointers): Likewise.
1786 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1788         PR ipa/86395
1789         * doc/invoke.texi (-fdump-ipa-): Document the "-optimized",
1790         "-missed", "-note", and "-all" sub-options.
1791         * ipa-inline.c (caller_growth_limits): Port from fprintf to dump
1792         API.
1793         (can_early_inline_edge_p): Likewise.
1794         (want_early_inline_function_p): Likewise.
1795         (want_inline_self_recursive_call_p): Likewise.
1796         (recursive_inlining): Likewise.
1797         (inline_small_functions): Likewise.
1798         (flatten_function): Likewise.
1799         (ipa_inline): Likewise.
1800         (inline_always_inline_functions): Likewise.
1801         (early_inline_small_functions): Likewise.
1802         (early_inliner): Likewise.
1803         * tree-inline.c (expand_call_inline): Likewise.
1805 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1807         * pretty-print.c (pp_format): Handle %f.
1808         (selftest::test_pp_format): Add test of %f.
1809         * pretty-print.h (pp_double): New macro.
1811 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1813         * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro.
1814         * dumpfile.c (make_item_for_dump_cgraph_node): Move to before...
1815         (dump_pretty_printer::decode_format): Implement "%C" for
1816         cgraph_node *.
1817         (selftest::test_capture_of_dump_calls): Rename "where" to
1818         "stmt_loc".  Convert test_decl to a function decl and set its
1819         location.  Add a symbol_table_test RAII instance and a
1820         cgraph_node, using it to test "%C" and dump_symtab_node.
1822 2018-11-08  Eric Botcazou  <ebotcazou@adacore.com>
1824         PR middle-end/87916
1825         * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P.
1827 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1829         * cgraph.c: Include "selftest.h".
1830         (saved_symtab): New variable.
1831         (selftest::symbol_table_test::symbol_table_test): New ctor.
1832         (selftest::symbol_table_test::~symbol_table_test): New dtor.
1833         (selftest::test_symbol_table_test): New test.
1834         (selftest::cgraph_c_tests): New.
1835         * cgraph.h (saved_symtab): New decl.
1836         (selftest::symbol_table_test): New class.
1837         * selftest-run-tests.c (selftest::run_tests): Call
1838         selftest::cgraph_c_tests.
1839         * selftest.h (selftest::cgraph_c_tests): New decl.
1841 2018-11-08  Richard Biener  <rguenther@suse.de>
1843         * tree-data-ref.h (lambda_int): New typedef.
1844         (lambda_vector_gcd): Adjust.
1845         (lambda_vector_new): Likewise.
1846         (lambda_matrix_new): Likewise.
1847         * tree-data-ref.c  (print_lambda_vector): Adjust.
1849 2018-11-08  Richard Biener  <rguenther@suse.de>
1851         PR tree-optimization/87929
1852         * tree-complex.c (expand_complex_comparison): Clean EH.
1854 2018-11-08  Martin Liska  <mliska@suse.cz>
1856         * doc/extend.texi: Reword.
1857         * predict.c (expr_expected_value_1): Likewise.
1859 2018-11-08  Richard Biener  <rguenther@suse.de>
1861         PR tree-optimization/87913
1862         * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares
1863         of extreme values to ordered comparisons.
1865 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1867         PR middle-end/42726
1868         * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior.
1870 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1872         * doc/invoke.texi: Remove leading dash from @opindex entries
1873         throughout the file.
1875 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1877         PR driver/80828
1878         * doc/invoke.texi (Option Summary): Add -e and --entry.
1879         (Link Options): Likewise.
1881 2018-11-07  Nathan Sidwell  <nathan@acm.org>
1883         PR 87926
1884         * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak
1885         --disable-checking bootstrap.
1887 2018-11-07  Hafiz Abid Qadeer  <abidh@codesourcery.com>
1889         * configure: Regenerated.
1891 2018-11-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1893         PR c/87691
1894         * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE
1895         to the mode of the widest field iff the widest field has mode class
1896         MODE_INT, or MODE_PARTIAL_INT and the union would be passed by
1897         reference.
1899 2018-11-07  Nikolai Merinov  <n.merinov@inango-systems.com>
1901         * common.opt: Add -Wattribute-warning.
1902         * doc/invoke.texi: Add documentation for -Wno-attribute-warning.
1903         * expr.c (expand_expr_real_1): Add new attribute to warning_at
1904         call to allow user configure behavior of "warning" attribute.
1906 2018-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
1908         * target.def: Put @: after every vs., e.g., and i.e. where it is
1909         followed by whitespace.
1910         * doc/extend.texi: Ditto.
1911         * doc/fragments.texi: Ditto.
1912         * doc/gimple.texi: Ditto.
1913         * doc/implement-c.texi: Ditto.
1914         * doc/install.texi: Ditto.
1915         * doc/invoke.texi: Ditto.
1916         * doc/md.texi: Ditto.
1917         * doc/plugins.texi: Ditto.
1918         * doc/rtl.texi: Ditto.
1919         * doc/sourcebuild.texi: Ditto.
1920         * doc/tm.texi.in: Ditto.
1921         * doc/ux.texi: Ditto.
1922         * doc/tm.texi: Regenerate.
1924 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1926         * config/arm/arm-cpus.in (ares): New entry.
1927         * config/arm/arm-tables.opt: Regenerate.
1928         * config/arm/arm-tune.md: Likewise.
1929         * doc/invoke.texi (ARM Options): Document ares.
1931 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1933         * config/aarch64/aarch64-cores.def (ares): Define.
1934         * config/aarch64/aarch64-tune.md: Regenerate.
1935         * doc/invoke.texi (AArch64 Options): Document ares value for mtune.
1937 2018-11-07  Jan Hubicka  <jh@suse.cz>
1939         * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE.
1940         (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and
1941         functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for
1942         TYPE_DECL.
1944 2018-11-07  Richard Biener  <rguenther@suse.de>
1946         PR tree-optimization/87914
1947         * tree-vect-loop.c (vect_is_simple_reduction): Improve detection
1948         of nested cycles.
1949         (vectorizable_reduction): Handle shifts and rotates by dispatching
1950         to vectorizable_shift.
1951         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle
1952         in-loop uses of vect_nested_cycle defs.  Merge cycle and internal
1953         def cases.
1954         (vectorizable_shift): Export and handle being called as
1955         vect_nested_cycle.
1956         (vect_analyze_stmt): Call vectorizable_shift after
1957         vectorizable_reduction.
1958         * tree-vectorizer.h (vectorizable_shift): Declare.
1960 2018-11-07  Jan Hubicka  <jh@suse.cz>
1962         * ipa-devirt.c (odr_types_equivalent_p): Expect constants
1963         than const decls in TREE_VALUE of enum.
1964         (dump_type_inheritance_graph): Improve duplicate dumping.
1965         (free_enum_values): New.
1966         (build_type_inheritance_graph): Use it.
1967         * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums
1968         which are not main variants or not ODR types.
1969         (verify_type_variant): Expect variants to have no TYPE_VALUES.
1971 2018-11-07  Richard Biener  <rguenther@suse.de>
1973         * ipa-inline.c (want_inline_small_function_p): Compute
1974         big_speedup_p lazily and last.
1976 2018-11-07  Jan Hubicka  <jh@suse.cz>
1978         * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when
1979         building incomplete variant of complete type.
1980         (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete
1981         variant of complete type.
1983 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
1985         * config/mips/mips.c: Fix typo in documentation of
1986         mips_loongson_ext2_prefetch_cookie.
1987         (mips_option_override): fix brain twister logical.
1988         * config/mips/mips.h: Fix typo in documentation of
1989         ISA_HAS_CTZ_CTO and define pattern.
1990         * config/mips/mips.md (prefetch): Hoist EXT2 above
1991         the 2EF/EXT block.
1992         (prefetch_indexed): Hoist EXT2 above the EXT block.
1994 2018-11-07  Jan Hubicka  <jh@suse.cz>
1996         * tree.c (free_lang_data_in_type): Add fld parameter; simplify
1997         return and parameter types of function and method types.
1998         (free_lang_data_in_cgraph): Update.
2000 2018-11-07  Martin Liska  <mliska@suse.cz>
2002         PR rtl-optimization/87868
2003         * postreload-gcse.c (eliminate_partially_redundant_load): Set
2004         threshold to max_count if we would overflow.
2005         * profile-count.h: Make max_count a public constant.
2007 2018-11-07  Martin Liska  <mliska@suse.cz>
2009         * mem-stats.h: Fix GNU coding style.
2011 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2013         * config/mips/gs264e.md: New.
2014         * config/mips/mips-cpus.def: Define gs264e.
2015         * config/mips/mips-tables.opt: Regenerate.
2016         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
2017         gs264e.
2018         (mips_issue_rate): Add support for gs264e.
2019         (mips_multipass_dfa_lookahead): Likewise.
2020         * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E.
2021         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e.
2022         (MIPS_ASE_MSA_SPEC): New.
2023         (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa.
2024         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E.
2025         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E.
2026         * config/mips/mips.md: Include gs264e.md.
2027         (processor): Add gs264e.
2028         * config/mips/mips.opt (MSA): Use Mask instead of Var.
2029         * doc/invoke.texi: Add gs264e to supported architectures.
2031 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2033         * config/mips/gs464e.md: New.
2034         * config/mips/mips-cpus.def: Define gs464e.
2035         * config/mips/mips-tables.opt: Regenerate.
2036         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
2037         gs464e.
2038         (mips_issue_rate): Add support for gs464e.
2039         (mips_multipass_dfa_lookahead): Likewise.
2040         (mips_option_override): Enable MMI, EXT and EXT2 for gs464e.
2041         * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E.
2042         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e.
2043         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E.
2044         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E.
2045         * config/mips/mips.md: Include gs464e.md.
2046         (processor): Add gs464e.
2047         * doc/invoke.texi: Add gs464e to supported architectures.
2049 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2051         * config/mips/loongson3a.md: Rename to ...
2052         * config/mips/gs464.md: ... here.
2053         * config/mips/mips-cpus.def: Define gs464; Add loongson3a
2054         as an alias of gs464 processor.
2055         * config/mips/mips-tables.opt: Regenerate.
2056         * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464
2057         instead of PROCESSOR_LOONGSON_3A.
2058         (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of
2059         TUNE_LOONGSON_3A.
2060         (mips_option_override): Enable MMI and EXT for gs464.
2061         * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464;
2062         Rename TUNE_LOONGSON_3A to TUNE_GS464.
2063         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464.
2064         (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4,
2065         ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of
2066         TARGET_LOONGSON_3A.
2067         * config/mips/mips.md: Include gs464.md instead of loongson3a.md.
2068         (processor): Add gs464;
2069         * doc/invoke.texi: Add gs464 to supported architectures.
2071 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2073         * config/mips/mips-protos.h
2074         (mips_loongson_ext2_prefetch_cookie): New prototype.
2075         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New.
2076         (mips_option_override): Enable TARGET_LOONGSON_EXT when
2077         TARGET_LOONGSON_EXT2 is true.
2078         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
2079         __mips_loongson_ext2, __mips_loongson_ext_rev=2.
2080         (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2.
2081         (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and
2082         TARGET_LOONGSON_EXT2.
2083         (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2.
2084         (define_insn "ctz<mode>2"): New insn pattern.
2085         (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2.
2086         (define_insn "prefetch_indexed_<mode>"): Include
2087         TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2.
2088         * config/mips/mips.opt (-mloongson-ext2): Add option.
2089         * gcc/doc/invoke.texi (-mloongson-ext2): Document.
2091 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2093         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add
2094         __mips_loongson_ext.
2095         (MIPS_ASE_LOONGSON_EXT_SPEC): New.
2096         (BASE_DRIVER_SELF_SPECS): march=loongson3a implies
2097         -mloongson-ext.
2098         (ASM_SPEC): Add mloongson-ext and mno-loongson-ext.
2099         * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo,
2100         <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT
2101         instead of TARGET_LOONGSON_3A.
2102         * config/mips/mips.opt (-mloongson-ext): Add option.
2103         * gcc/doc/invoke.texi (-mloongson-ext): Document.
2105 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2107         * config.gcc (extra_headers): Add loongson-mmiintrin.h.
2108         * config/mips/loongson.md: Move to ...
2109         * config/mips/loongson-mmi.md: here; Adjustment.
2110         * config/mips/loongson.h: Move to ...
2111         State as deprecated. Include loongson-mmiintrin.h for back
2112         compatibility and warning.
2113         * config/mips/loongson-mmiintrin.h: ... here.
2114         * config/mips/mips.c (mips_hard_regno_mode_ok_uncached,
2115         mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use
2116         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
2117         (mips_option_override): Make sure MMI use hard float;
2118         (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd,
2119         mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast,
2120         mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of
2121         TARGET_LOONGSON_VECTORS.
2122         * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete.
2123         (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi.
2124         (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New.
2125         (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies
2126         -mloongson-mmi.
2127         (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of
2128         TARGET_LOONGSON_VECTORS.
2129         * gcc/config/mips/mips.md (MOVE64, MOVE128): Use
2130         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
2131         (Loongson MMI patterns): Include loongson-mmi.md instead of
2132         loongson.md.
2133         * gcc/config/mips/mips.opt (-mloongson-mmi): New option.
2134         * gcc/doc/invoke.texi (-mloongson-mmi): Document.
2136 2018-11-07  Richard Biener  <rguenther@suse.de>
2138         PR lto/87906
2139         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
2140         BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
2142 2018-11-07  Alexandre Oliva <aoliva@redhat.com>
2144         PR rtl-optimization/87874
2145         * lra.c (lra_substitute_pseudo): Do not create a subreg for
2146         const wide ints.
2148 2018-11-06  Aaron Sawdey  <acsawdey@linux.ibm.com>
2150         * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
2151         if not in indexed or indirect form.
2152         (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
2153         (bswap<mode>2_store): Ditto.
2155 2018-11-06  Richard Earnshaw  <rearnsha@arm.com>
2157         * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
2158         the UNSPEC.
2160 2018-11-06  Richard Biener  <rguenther@suse.de>
2162         PR tree-optimization/86850
2163         * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
2164         instead of src.m_vec.
2166 2018-11-06  Jan Hubicka  <jh@suse.cz>
2168         * tree.c (fld_simplified_type_name): Break out form ...
2169         (free_lang_data_in_type): ... here.
2170         (fld_type_variant_equal_p): Use it.
2172 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2174         * config/default-d.c: Include memmodel.h.
2176         * config/sol2-d.c: New file.
2177         * config/t-sol2 (sol2-d.o): New rule.
2178         * config.gcc <*-*-solaris2*>: Set d_target_objs,
2179         target_has_targetdm.
2181 2018-11-06  Jan Hubicka  <jh@suse.cz>
2183         * tree.c (fld_type_variant): Also copy alignment; be sure that
2184         new variant is equal.
2186 2018-11-06  Ilya Leoshkevich  <iii@linux.ibm.com>
2188         PR target/87762
2189         * config/s390/s390.md: Add relative_long attribute.
2191 2018-11-06  Jan Hubicka  <jh@suse.cz>
2193         * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
2194         * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
2195         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
2196         stream TYPE_NEEDS_CONSTRUCTING.
2197         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
2198         * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
2200 2018-11-06  Richard Biener  <rguenther@suse.de>
2202         * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
2203         dump-scope ...
2204         (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
2206 2018-11-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
2208         * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
2209         REG_CLASS_CONTENTS[GEN_REGS].
2210         (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
2212 2018-11-06  Jan Hubicka  <jh@suse.cz>
2214          * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
2216 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2218         PR sanitizer/80953
2219         * config/sol2.h (ASAN_CC1_SPEC): Define.
2220         (LD_WHOLE_ARCHIVE_OPTION): Define.
2221         (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
2222         (ASAN_REJECT_SPEC): Provide default.
2223         (LIBASAN_EARLY_SPEC): Define.
2224         (LIBTSAN_EARLY_SPEC): Define.
2225         (LIBLSAN_EARLY_SPEC): Define.
2226         * config/i386/sol2.h (CC1_SPEC): Redefine.
2227         (ASAN_REJECT_SPEC): Define.
2229         * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
2230         (TARGET_ASAN_SHADOW_OFFSET): Define.
2231         (sparc_asan_shadow_offset): New function.
2232         * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
2233         (ASAN_REJECT_SPEC): Define.
2235 2018-11-06  Jan Hubicka  <jh@suse.cz>
2237         * tree.c (fld_type_variant): Copy canonical type.
2238         (fld_incomplete_type_of): Check that canonical types looks sane;
2239         copy canonical type.
2240         (verify_type): Accept when incomplete type has complete canonical type.
2242 2018-11-06  Jan Hubicka  <jh@suse.cz>
2244         * tree.c (free_lang_data): Reset overwite_assembler_name,
2245         print_xnode, print_decl, print_type and print_identifier of
2246         langhooks.
2248 2018-11-06  Richard Biener  <rguenther@suse.de>
2250         PR tree-optimization/87889
2251         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
2252         Do nothing if old and new arg are the same
2254 2018-11-06  Andreas Krebbel  <krebbel@linux.ibm.com>
2256         PR target/87723
2257         * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
2258         attributes for operands 3 and 4.
2260 2018-11-06  Richard Biener  <rguenther@suse.de>
2262         PR middle-end/18041
2263         * simplify-rtx.c (simplify_binary_operation_1): Add pattern
2264         matching bitfield insertion.
2266 2018-11-06  Alexandre Oliva <aoliva@redhat.com>
2268         * auto-inc-dec.c: Include valtrack.h.  Improve comments.
2269         (reg_next_debug_use): New.
2270         (attempt_change): Propagate adjusted expression into affected
2271         debug insns.
2272         (merge_in_block): Track uses in debug insns.
2273         (pass_inc_dec::execute): Allocate and release
2274         reg_next_debug_use.
2276 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
2278         * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
2279         (_mm512_fixupimm_round_pd): Update parameters and builtin.
2280         (_mm512_maskz_fixupimm_round_pd): Ditto.
2281         (_mm512_fixupimm_round_ps): Ditto.
2282         (_mm512_maskz_fixupimm_round_ps): Ditto.
2283         (_mm_fixupimm_round_sd): Ditto.
2284         (_mm_maskz_fixupimm_round_sd): Ditto.
2285         (_mm_fixupimm_round_ss): Ditto.
2286         (_mm_maskz_fixupimm_round_ss): Ditto.
2287         (_mm512_fixupimm_pd): Ditto.
2288         (_mm512_maskz_fixupimm_pd): Ditto.
2289         (_mm512_fixupimm_ps): Ditto.
2290         (_mm512_maskz_fixupimm_ps): Ditto.
2291         (_mm_fixupimm_sd): Ditto.
2292         (_mm_maskz_fixupimm_sd): Ditto.
2293         (_mm_fixupimm_ss): Ditto.
2294         (_mm_maskz_fixupimm_ss): Ditto.
2295         (_mm512_mask_fixupimm_round_pd): Update builtin.
2296         (_mm512_mask_fixupimm_round_ps): Ditto.
2297         (_mm_mask_fixupimm_round_sd): Ditto.
2298         (_mm_mask_fixupimm_round_ss): Ditto.
2299         (_mm512_mask_fixupimm_pd): Ditto.
2300         (_mm512_mask_fixupimm_ps): Ditto.
2301         (_mm_mask_fixupimm_sd): Ditto.
2302         (_mm_mask_fixupimm_ss): Ditto.
2303         * config/i386/avx512vlintrin.h:
2304         (_mm256_fixupimm_pd): Update parameters and builtin.
2305         (_mm256_maskz_fixupimm_pd): Ditto.
2306         (_mm256_fixupimm_ps): Ditto.
2307         (_mm256_maskz_fixupimm_ps): Ditto.
2308         (_mm_fixupimm_pd): Ditto.
2309         (_mm_maskz_fixupimm_pd): Ditto.
2310         (_mm_fixupimm_ps): Ditto.
2311         (_mm_maskz_fixupimm_ps): Ditto.
2312         (_mm256_mask_fixupimm_pd): Update builtin.
2313         (_mm256_mask_fixupimm_ps): Ditto.
2314         (_mm_mask_fixupimm_pd): Ditto.
2315         (_mm_mask_fixupimm_ps): Ditto.
2316         * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
2317         * config/i386/i386-builtin.def: Update builtin definitions.
2318         * config/i386/i386.c: Handle new builtin types and remove useless ones.
2319         * config/i386/sse.md: Update VFIXUPIMM* patterns.
2320         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
2321         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
2322         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
2323         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
2324         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
2325         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
2326         * config/i386/subst.md:
2327         (round_saeonly_sd_mask_operand4): Add new subst_attr.
2328         (round_saeonly_sd_mask_op4): Ditto.
2329         (round_saeonly_expand_operand5): Ditto.
2330         (round_saeonly_expand): Update.
2332 2018-11-05  Max Filippov  <jcmvbkbc@gmail.com>
2334         * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
2336 2018-11-05  Segher Boessenkool  <segher@kernel.crashing.org>
2338         PR rtl-optimization/87871
2339         * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
2341 2018-11-05  Paul Koning  <ni1d@arrl.net>
2343         * doc/sourcebuild.texi (target attributes): Document new "inf"
2344         effective target keyword.
2346 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
2348         * config/s390/s390.c (s390_register_move_cost): Increase costs for
2349         moves involving the CC reg.
2351 2018-11-05  Richard Biener  <rguenther@suse.de>
2353         PR tree-optimization/87873
2354         * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
2355         argument.
2356         * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
2357         * tree-vect-loop.c (vect_transform_loop): When splitting the
2358         loop exit also create forwarder PHIs for constants.
2359         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
2360         Handle constant to_arg, add extra checking we match up the correct
2361         PHIs.
2363 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
2365         * config/s390/s390.md: QImode and HImode for load on condition.
2367 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
2369         * config/s390/predicates.md: Fix typo.
2370         * config/s390/s390.md: Allow immediates for load on condition.
2372 2018-11-05  Martin Liska  <mliska@suse.cz>
2374         * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
2375         * bitmap.h (struct bitmap_usage): Likewise.
2376         * ggc-common.c (SCALE): Remove.
2377         (LABEL): Likewise.
2378         (struct ggc_usage): Use SIZE_AMOUNT. And update
2379         compare method.
2380         * ggc-page.c (SCALE): Remove.
2381         (STAT_LABEL): Remove.
2382         (ggc_print_statistics): Use SIZE_AMOUNT.
2383         * gimple.h (SCALE): Remove.
2384         (LABEL): Likewise.
2385         * input.c (ONE_K): Remove.
2386         (ONE_M): Likewise.
2387         (SCALE): Likewise.
2388         (STAT_LABEL): Likewise.
2389         (FORMAT_AMOUNT): Likewise.
2390         (dump_line_table_statistics): Use SIZE_AMOUNT.
2391         * mem-stats.h (struct mem_usage): Likewise.
2392         * rtl.c (dump_rtx_statistics): Likewise.
2393         (rtx_alloc_counts): Change type to size_t.
2394         (rtx_alloc_sizes): Likewise.
2395         (rtx_count_cmp): New.
2396         (dump_rtx_statistics): Sort first based on counts.
2397         * tree.c (tree_nodes_cmp): New.
2398         (tree_codes_cmp): New.
2399         (dump_tree_statistics): Sort first based on counts.
2400         * system.h (ONE_K): New.
2401         (ONE_M): Likewise.
2402         (SIZE_SCALE): Likewise.
2403         (SIZE_LABEL): Likewise.
2404         (SIZE_AMOUNT): Likewise.
2405         * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
2406         * tree-dfa.c (dump_dfa_stats): Likewise.
2407         * tree-phinodes.c (phinodes_print_statistics): Likewise.
2408         * tree-ssanames.c (ssanames_print_statistics): Likewise.
2409         * tree.c (dump_tree_statistics): Likewise.
2410         * vec.c (struct vec_usage): Likewise.
2411         * trans-mem.c (tm_mangle): Enlarge buffer in order to not
2412         trigger a -Werror=format-overflow with
2413         --enable-gather-detailed-stats.
2415 2018-11-05  Martin Liska  <mliska@suse.cz>
2417         * mem-stats.h (mem_alloc_description::release_instance_overhead):
2418         Return T *.
2419         * vec.c (struct vec_usage): Register m_element_size.
2420         (vec_prefix::register_overhead): New arguments: elements and
2421         element_size.
2422         (vec_prefix::release_overhead): Subtract elements.
2423         * vec.h (struct vec_prefix): Change signature.
2424         (va_heap::reserve): Pass proper arguments.
2425         (va_heap::release): Likewise.
2427 2018-11-05  Martin Liska  <mliska@suse.cz>
2429         * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
2430         style.
2431         * vec.c: Likewise.
2433 2018-11-05  Richard Biener  <rguenther@suse.de>
2435         * tree-scalar-evolution.h (final_value_replacement_loop): Update
2436         prototype.
2437         * tree-scalar-evolution.c (final_value_replacement_loop): Return
2438         whether anything was done.
2439         (scev_const_prop): Remove constant propagation part, fold
2440         remains into ...
2441         * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
2442         (pass_data_scev_cprop): TODO_cleanup_cfg is now done
2443         conditionally.
2445 2018-11-05  Jakub Jelinek  <jakub@redhat.com>
2447         PR tree-optimization/87859
2448         * gimple-ssa-store-merging.c (struct merged_store_group): Add
2449         only_constants and first_nonmergeable_order members.
2450         (merged_store_group::merged_store_group): Initialize them.
2451         (merged_store_group::do_merge): Clear only_constants member if
2452         adding something other than INTEGER_CST store.
2453         (imm_store_chain_info::coalesce_immediate_stores): Don't merge
2454         stores with order >= first_nonmergeable_order.  Use
2455         merged_store->only_constants instead of always recomputing it.
2456         Set merged_store->first_nonmergeable_order if we've skipped any
2457         stores.  Attempt to merge overlapping INTEGER_CST stores that
2458         we would otherwise skip.
2460         PR sanitizer/87837
2461         * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
2463 2018-11-05  Xuepeng Guo  <xuepeng.guo@intel.com>
2465         PR target/87853
2466         * config/i386/emmintrin.h (__v16qs): New to cope with option
2467         -funsigned-char.
2468         (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
2469         (_mm_cmplt_epi8): Likewise.
2470         (_mm_cmpgt_epi8): Likewise.
2472 2018-11-05  Richard Biener  <rguenther@suse.de>
2474         PR rtl-optimization/87852
2475         * fwprop.c (use_killed_between): Only consider single-defs of the
2476         use whose definition statement dominates the use.
2478 2018-11-05  Martin Liska  <mliska@suse.cz>
2480         PR web/87829
2481         * doc/invoke.texi: Remove options that are
2482         not disabled with -Os.
2484 2018-11-05  Martin Liska  <mliska@suse.cz>
2486         PR c/87811
2487         * doc/extend.texi: Update constrain about the last argument
2488         of __builtin_expect_with_probability.
2490 2018-11-05  Martin Liska  <mliska@suse.cz>
2492         PR c/87811
2493         * predict.c (expr_expected_value_1): Verify
2494         that last argument is a real constants and emit
2495         error.
2497 2018-11-05  Martin Liska  <mliska@suse.cz>
2499         PR gcov-profile/77698
2500         * ipa-profile.c (ipa_profile): Adjust hotness threshold
2501         only in LTO mode.
2503 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2505         PR tree-optimization/86572
2506         * builtins.c (c_strlen): Handle negative offsets in a safe way.
2508 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2510         PR tree-optimization/87672
2511         * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
2512         * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
2514 2018-11-04  Uros Bizjak  <ubizjak@gmail.com>
2516         PR middle-end/58372
2517         * cfgexpand.c (pass_expand::execute): Move the call to
2518         finish_eh_generation in front of the call to expand_stack_alignment.
2520 2018-11-04  Venkataramanan Kumar <venkataramanan.kumar@amd.com>
2522         * common/config/i386/i386-common.c (processor_alias_table): Add
2523         znver2 entry.
2524         * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
2525         (case ${target}): Add znver2.
2526         * config/i386/driver-i386.c: (host_detect_local_cpu): Let
2527         -march=native recognize znver2 processors.
2528         * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
2529         * config/i386/i386.c (m_znver2): New definition.
2530         (m_ZNVER): New definition.
2531         (m_AMD_MULTIPLE): Includes m_znver2.
2532         (processor_cost_table): Add znver2 entry.
2533         (processor_target_table): Add znver2 entry.
2534         (get_builtin_code_for_version): Set priority for
2535         PROCESSOR_ZNVER2.
2536         (processor_model): Add M_AMDFAM17H_ZNVER2.
2537         (arch_names_table): Ditto.
2538         (ix86_reassociation_width): Include znver2.
2539         * config/i386/i386.h (TARGET_znver2): New definition.
2540         (struct ix86_size_cost): Add TARGET_ZNVER2.
2541         (enum processor_type): Add PROCESSOR_ZNVER2.
2542         * config/i386/i386.md (define_attr "cpu"): Add znver2.
2543         * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
2544         * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
2545         (ix86_adjust_cost): Add znver2.
2546         * config/i386/x86-tune.def:  Replace m_ZNVER1 by m_ZNVER.
2547         * gcc/doc/extend.texi: Add details about znver2.
2548         * gcc/doc/invoke.texi: Add details about znver2.
2550 2018-11-03  Sandra Loosemore  <sandra@codesourcery.com>
2552         PR target/87079
2554         * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
2555         pattern.
2557 2018-11-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
2559         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
2560         attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
2562 2018-11-02  Richard Earnshaw  <rearnsha@arm.com>
2564         * config/aarch64/aarch64.c ((aarch64_override_options): Disable
2565         shrink-wrapping when -mtrack-speculation.
2567 2018-11-02  Richard Biener  <rguenther@suse.de>
2569         * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
2570         (pop_cost_one_pair): Do not free pair.
2571         (pop_best_coalesce): Likewise.
2572         (create_coalesce_list): Initialize obstack.
2573         (delete_coalesce_list): Free obstack.
2574         (find_coalesce_pair): Obstack-allocate coalesce pairs.
2575         (add_cost_one_coalesce): Likewise.
2576         (struct live_track): Remove bitmap pointer indirections.
2577         (new_live_track): Adjust.
2578         (delete_live_track): Likewise.
2579         (live_track_remove_partition): Likewise.
2580         (live_track_add_partition): Likewise.
2581         (live_track_live_p): Likewise.
2582         (live_track_process_def): Likewise.
2583         (live_track_clear_base_vars): Likewise.
2585 2018-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2587         * configure.ac (gcc_cv_as_sparc_register_op): Remove.
2588         * configure: Regenerate.
2589         * config.in: Regenerate.
2590         * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
2591         HAVE_AS_REGISTER_PSEUDO_OP guard.
2592         * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
2593         !HAVE_AS_REGISTER_PSEUDO_OP support.
2594         (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
2595         guard.
2597 2018-11-02  Richard Biener  <rguenther@suse.de>
2599         * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
2600         with is_type_die.
2602 2018-11-02  Richard Biener  <rguenther@suse.de>
2604         PR tree-optimization/87776
2605         * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
2606         executable when iterating but running into rpo-vn-max-loop-depth
2607         and not eliding the iteration.
2609 2018-11-30  Jan Hubicka  <jh@suse.cz>
2611         * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
2612         TYPE_DECL.
2614 2018-11-01  Aaron Sawdey  <acsawdey@linux.ibm.com>
2616         * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
2617         prototype.
2618         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
2619         Combine with rs6000_address_for_fpconvert.
2620         (rs6000_address_for_fpconvert): Combine with
2621         rs6000_force_indexed_or_indirect_mem.
2622         (rs6000_expand_vector_init): Change function call from
2623         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
2624         * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
2625         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
2626         (floatsi<mode>2_lfiwax_mem): Ditto.
2627         (floatunssi<mode>2_lfiwzx): Ditto.
2628         (floatunssi<mode>2_lfiwzx_mem): Ditto.
2629         (float<QHI:mode><FP_ISA3:mode>2): Ditto.
2630         (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
2631         (fix_trunc<mode>si2_stfiwx): Ditto.
2632         (fixuns_trunc<mode>si2_stfiwx): Ditto.
2633         (float_<mode>si2_hw): Ditto.
2634         (floatuns_<mode>si2_hw): Ditto.
2635         * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
2636         (vsx_splat_<mode>): Ditto.
2638 2018-11-01  Joseph Myers  <joseph@codesourcery.com>
2640         * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
2641         at end of assembler input text.
2642         * configure: Regenerate.
2644 2018-11-01  Jakub Jelinek  <jakub@redhat.com>
2646         PR tree-optimization/87826
2647         * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
2648         negative or larger or equal to type's precision.
2650 2018-10-31  Alexandre Oliva <aoliva@redhat.com>
2652         * opts.c (default_options_table): Do not enable
2653         OPT_fdelayed_branch at -Og.
2654         * doc/invoke.texi (-fdelayed-branch): Document it.
2656 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
2658         * optabs-libfuncs.c (build_libfunc_function_visibility):
2659         New, split out from...
2660         (build_libfunc_function): ... here.
2661         (init_one_libfunc_visibility): New, split out from ...
2662         (init_one_libfunc): ... here.
2664         * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
2665         scratch register need not be early-clobber.  Document the reason
2666         why we cannot use ST<OP>.
2668 2018-10-31  Joseph Myers  <joseph@codesourcery.com>
2670         PR bootstrap/82856
2671         * configure.ac: Remove AC_PREREQ.  Use AC_LANG_SOURCE.  Use single
2672         line for second argument of AC_DEFINE_UNQUOTED.
2673         * doc/install.texi (Tools/packages necessary for modifying GCC):
2674         Update to autoconf 2.69 and automake 1.15.1.
2675         * aclocal.m4, config.in, configure: Regenerate.
2677 2018-10-31  Pat Haugen  <pthaugen@us.ibm.com>
2679         * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
2680         initialization.
2682 2018-10-31  Martin Liska  <mliska@suse.cz>
2684         PR driver/83193
2685         * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
2686         Add new argument invalid_extension.
2687         (aarch64_get_all_extension_candidates): New function.
2688         (aarch64_rewrite_selected_cpu): Add NULL to function call.
2689         * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
2690         new argument.
2691         (aarch64_get_all_extension_candidates): New function.
2692         * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
2693         argument invalid_extension.
2694         (aarch64_parse_cpu): Likewise.
2695         (aarch64_print_hint_for_extensions): New function.
2696         (aarch64_validate_mcpu): Provide hint about invalid extension.
2697         (aarch64_validate_march): Likewise.
2698         (aarch64_handle_attr_arch): Pass new argument.
2699         (aarch64_handle_attr_cpu): Provide hint about invalid extension.
2700         (aarch64_handle_attr_isa_flags): Likewise.
2702 2018-10-31  Richard Biener  <rguenther@suse.de>
2704         PR middle-end/70359
2705         PR middle-end/86270
2706         * tree-outof-ssa.c (insert_backedge_copies): Restrict
2707         copy generation to useful cases.  Place the copy before
2708         the definition of the backedge value when possible.
2710 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2712         * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
2713         * config/arc/arc.c (arc_active_insn): New function.
2714         (check_store_cacheline_hazard): Likewise.
2715         (workaround_arc_anomaly): Use check_store_cacheline_hazard.
2716         (arc_override_options): Disable delay slot scheduler for older
2717         A7.
2718         (arc_store_addr_hazard_p): New implementation, old one renamed to
2719         ...
2720         (arc_store_addr_hazard_internal_p): Renamed.
2721         (arc_reorg): Don't combine into brcc instructions which are part
2722         of hardware hazard solution.
2723         * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
2724         (tune_arc700): Likewise.
2725         * config/arc/arc.opt (arc7xx): New tune value.
2726         * config/arc/arc700.md: Improve A7 scheduler.
2728 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2730         * config/arc/arc.c (arc_override_options): Remove
2731         TARGET_COMPACT_CASESI.
2732         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
2733         (CASE_VECTOR_MODE): Likewise.
2734         (CASE_VECTOR_PC_RELATIVE): Likewise.
2735         (CASE_VECTOR_SHORTEN_MODE): Likewise.
2736         (CASE_VECTOR_SHORTEN_MODE1): Delete.
2737         (ADDR_VEC_ALIGN): Update.
2738         (ASM_OUTPUT_CASE_LABEL): Undefine.
2739         (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
2740         (TARGET_BI_BIH): Define.
2741         (DEFAULT_BRANCH_INDEX): Likewise.
2742         * config/arc/arc.md (casesi): Rework to accept BI/BIH
2743         instructions, remove compact_casesi use case.
2744         (casesi_compact_jump): Remove.
2745         (casesi_dispatch): New pattern.
2746         * config/arc/arc.opt: Add mbranch-index option. Deprecate
2747         compact_casesi option.
2748         * doc/invoke.texi: Document mbranch-index option.
2750 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2752         * config/arc/arc.c (arc_get_tp): Remove function.
2753         (arc_emit_call_tls_get_addr): Likewise.
2754         (arc_call_tls_get_addr): New function.
2755         (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
2756         * config/arc/arc.md (tls_load_tp_soft): Remove.
2757         (tls_gd_get_addr): Likewise.
2759 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2761         * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
2762         (mulsi3_600_lib): Remove pattern.
2763         (umulsi3_highpart_600_lib_le): Likewise.
2764         (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
2765         (umulsidi3): Remove call to umulsidi3_600_lib.
2766         (umulsidi3_600_lib): Remove pattern.
2767         (peephole2): Remove peephole using the above deprecated patterns.
2769 2018-10-31  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
2771         PR target/87374
2772         * config/arm/arm.c (arm_option_check_internal): Disable the combined
2773         use of -mslow-flash-data and -mword-relocations.
2774         (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
2775         * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
2776         flag_pic.
2777         * doc/invoke.texi (-mword-relocations): Mention conflict with
2778         -mslow-flash-data.
2779         (-mslow-flash-data): Reciprocally.
2781 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
2783         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
2784         16-byte modes held in GP registers to use an even regno.
2786         * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
2787         (aarch64_atomic_ldop_supported_p): Remove.
2788         (aarch64_gen_atomic_ldop): Remove.
2789         * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
2790         Fully expand LSE operations here.
2791         (atomic_fetch_<atomic_optab><ALLI>): Likewise.
2792         (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
2793         (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
2794         and use ATOMIC_LDOP instead; use register_operand for the input;
2795         drop the split and emit insns directly.
2796         (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
2797         (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
2798         (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
2800         * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
2801         (aarch64_gen_atomic_ldop): Don't call it.
2802         * config/aarch64/atomics.md (atomic_exchange<ALLI>):
2803         Use aarch64_reg_or_zero.
2804         (aarch64_atomic_exchange<ALLI>): Likewise.
2805         (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
2806         operand 0; use aarch64_reg_or_zero for input; merge ...
2807         (@aarch64_atomic_swp<ALLI>): ... this and remove.
2809         * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
2810         (aarch64_split_compare_and_swap): Use it.
2811         (aarch64_expand_compare_and_swap): Likewise.  Remove convert_modes;
2812         test oldval against the proper predicate.
2813         * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
2814         Use nonmemory_operand for expected.
2815         (cas_short_expected_pred): New.
2816         (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
2817         (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
2818         * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
2819         (aarch64_plushi_operand): New.
2821         * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
2822         Force oldval into the rval register for TARGET_LSE; emit the compare
2823         during initial expansion so that it may be deleted if unused.
2824         (aarch64_gen_atomic_cas): Remove.
2825         * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
2826         Change =&r to +r for operand 0; use match_dup for operand 2;
2827         remove is_weak and mod_f operands as unused.  Drop the split
2828         and merge with...
2829         (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
2830         (@aarch64_compare_and_swap<GPI>_lse): Similarly.
2831         (@aarch64_atomic_cas<GPI>): Similarly.
2833 2018-10-31  Richard Biener  <rguenther@suse.de>
2835         * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
2836         using ABSU_EXPR.
2838 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2840         * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
2841         * config/aarch64/aarch64.md: Include saphira.md
2842         * config/aarch64/saphira.md: New file for pipeline description.
2844 2018-10-30  Martin Sebor  <msebor@redhat.com>
2846         PR middle-end/87041
2847         * gimple-ssa-sprintf.c (format_directive): Use %G to include
2848         inlining context.
2849         (sprintf_dom_walker::compute_format_length):
2850         Avoid setting POSUNDER4K here.
2851         (get_destination_size): Handle null argument values.
2852         (get_user_idx_format): New function.
2853         (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
2854         functions, including user-defined with attribute format printf.
2855         Use %G to include inlining context.
2856         Set POSUNDER4K here.
2858 2018-10-30  Jan Hubicka  <jh@suse.cz>
2860         * params.def (lto-partitions): Bump from 32 to 128.
2862 2018-10-30  Jan Hubicka  <jh@suse.cz>
2864         * tree.c
2865         (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
2866         head in file.
2867         (free_lang_data_in_type): Forward declare.
2868         (fld_type_variant_equal_p): New function.
2869         (fld_type_variant): New function
2870         (fld_incomplete_types): New hash.
2871         (fld_incomplete_type_of): New function
2872         (fld_simplfied-type): New function.
2873         (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
2874         (free_lang_data): Allocate and free fld_incomplete_type; update call
2875         of free_lang_data_in_decl.
2877 2018-10-30  Eric Botcazou  <ebotcazou@adacore.com>
2879         * gcov.c (output_lines): Remove duplicate line.
2881 2018-10-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
2883         * config/rs6000/rs6000.md (bswapdi2): Force address into register
2884         if not in indexed or indirect form.
2885         (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
2886         (bswapdi2_store): Ditto.
2887         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
2888         helper function.
2889         * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
2890         Prototype for helper function.
2892 2018-10-30  Martin Sebor  <msebor@redhat.com>
2894         * doc/extend.texi (optimize): Clarify/expand attribute documentation.
2895         (target, pragma GCC optimize, pragma GCC target): Ditto.
2897 2018-10-30  Jonathan Wakely  <jwakely@redhat.com>
2899         * doc/extend.texi: Fix prototype and description of
2900         __builtin_expect_with_probability.
2902 2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
2904         * cgraph.h (clone_function_name_1): Replaced by new
2905           clone_function_name_numbered that takes name as string; for
2906           privatize_symbol_name_1 use only.
2907           (clone_function_name): Renamed to
2908           clone_function_name_numbered to be explicit about numbering.
2909           (clone_function_name): New two-argument function that does
2910           not number its output.
2911           (clone_function_name): New three-argument function that
2912           takes a number to append to its output.
2913         * cgraphclones.c (duplicate_thunk_for_node):
2914           (clone_function_name_1): Renamed.
2915           (clone_function_name_numbered): Two new functions.
2916           (clone_function_name): Improved documentation.
2917           (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
2918         * config/rs6000/rs6000.c (make_resolver_func): Ditto.
2919         * final.c (final_scan_insn_1): Use the new clone_function_name
2920           without numbering.
2921         * multiple_target.c (create_dispatcher_calls): Ditto.
2922           (create_target_clone): Ditto.
2923         * omp-expand.c (grid_expand_target_grid_body): Ditto.
2924         * omp-low.c (create_omp_child_function_name): Ditto.
2925         * omp-simd-clone.c (simd_clone_create): Ditto.
2926         * symtab.c (simd_symtab_node::noninterposable_alias): Use the
2927           new clone_function_name without numbering.
2929 2018-10-30  Richard Earnshaw  <rearnsha@arm.com>
2931         * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
2932         Assert that the allocation size is not zero.
2934 2018-10-30  Richard Biener  <rguenther@suse.de>
2936         PR tree-optimization/87800
2937         * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
2938         non-induction or reduction PHIs.
2940 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
2942         * config/aarch64/falkor-tag-collision-avoidance.c
2943         (execute_tag_collision_avoidance): Call df_note_add_problem.
2945 2018-10-30  Martin Liska  <mliska@suse.cz>
2947         * doc/extend.texi: Fix typo in documentation
2948         of __builtin_expect_with_probability.
2950 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
2952         PR c++/87721
2953         * input.c (get_substring_ranges_for_loc): Detect if
2954         linemap_resolve_location gives us a NULL map, and reject
2955         this case.
2957 2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>
2959         * config.gcc (xstormy16-*-elf): Set tm_d_file.
2961 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
2962             Martin Sebor  <msebor@redhat.com>
2963             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2965         * doc/ux.texi (Quoting): New subsection, adapted from material at
2966         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
2967         MartinSebor and ManuelLopezIbanez.
2968         (Fix-it hints): Note that fix-it hints shouldn't be marked for
2969         translation.
2971 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
2973         PR middle-end/87469
2974         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
2975         max value.
2977 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
2979         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
2981 2018-10-29  Paul Koning  <ni1d@arrl.net>
2983         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
2985 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
2987         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
2988         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
2989         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
2990         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
2991         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
2992         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
2993         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
2994         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
2995         Change 'vector' to '__vector'.
2996         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
2997         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
2998         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
2999         _mm_avg_pu16): Likewise.  And, whitespace corrections.
3001 2018-10-29  Richard Biener  <rguenther@suse.de>
3003         PR tree-optimization/87785
3004         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
3005         internal defs.
3007 2018-10-29  Olivier Hainque  <hainque@adacore.com>
3009         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
3011 2018-10-29  Olivier Hainque  <hainque@adacore.com>
3013         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
3014         ports configuration macro, defaults to "gnu".
3015         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
3016         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
3018 2018-10-29  Olivier Hainque  <hainque@adacore.com>
3020         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
3022 2018-10-29  Richard Biener  <rguenther@suse.de>
3024         PR tree-optimization/87790
3025         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
3026         (vect_make_slp_decision): Adjust.
3027         (vect_slp_analyze_bb_1): Likewise.
3028         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
3029         edges.
3031 2018-10-29  Richard Biener  <rguenther@suse.de>
3033         PR tree-optimization/87785
3034         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
3035         and processing.
3036         (vect_build_slp_tree): Likewise.
3037         (vect_gather_slp_loads): New function.
3038         (vect_analyze_slp_instance): Gather loads separately from the
3039         SLP tree build.
3041 2018-10-29  Martin Liska  <mliska@suse.cz>
3043         * Makefile.in: Make dependency to json.o.
3044         * doc/gcov.texi: Document new JSON format, remove
3045         old intermediate format documentation.
3046         * gcov.c (struct function_info): Come up with m_name and
3047         m_demangled_name.
3048         (function_info::function_info): Initialize it.
3049         (function_info::~function_info): Release it.
3050         (main): Rename flag_intermediate_format to flag_json_format.
3051         (print_usage): Describe --json-format.
3052         (process_args): Set flag_json_format.
3053         (output_intermediate_line): Remove.
3054         (output_intermediate_json_line): Likewise.
3055         (get_gcov_intermediate_filename): Return new extension
3056         ".gcov.json.gz".
3057         (output_intermediate_file): Implement JSON emission.
3058         (output_json_intermediate_file): Implement JSON emission.
3059         (generate_results): Use ::get_name for function name.
3060         Handle JSON output file.
3061         (read_graph_file): Use ::get_name instead of cplus_demangle.
3062         (read_count_file): Likewise.
3063         (solve_flow_graph): Likewise.
3064         (add_line_counts): Likewise.
3065         (accumulate_line_counts): Use new flag_json_format.
3066         (output_function_details): Use ::get_name instead of cplus_demangle.
3067         (output_lines): Likewise.
3068         * json.cc (test_writing_literals): Add new tests.
3069         * json.h (class literal): Add new boolean constructor.
3071 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
3073         PR rtl-optimization/87701
3074         PR rtl-optimization/87780
3075         * combine.c (make_more_copies): Rewrite.
3077 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
3079         * doc/generic.texi (ABSU_EXPR): Document.
3080         * match.pd (absu(x)*absu(x) -> x*x): Handle.
3081         (absu(absu(X)) -> absu(X)): Likewise.
3082         (absu(-X) -> absu(X)): Likewise.
3083         (absu(X)  where X is nonnegative -> X): Likewise.
3085 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
3087         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
3088         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
3089         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
3090         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
3091         (s-tm-texi): Also check timestamp on d-target.def.
3092         (generated_files): Add TM_D_H and d-target-hooks-def.h.
3093         (build/genhooks.o): Also depend on D_TARGET_DEF.
3094         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
3095         variables.
3096         * config/aarch64/aarch64-d.c: New file.
3097         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
3098         Define.
3099         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
3100         prototype.
3101         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
3102         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
3103         * config/arm/arm-d.c: New file.
3104         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
3105         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
3106         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
3107         * config/arm/t-arm (arm-d.o): New rule.
3108         * config/default-d.c: New file.
3109         * config/glibc-d.c: New file.
3110         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3111         * config/i386/i386-d.c: New file.
3112         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
3113         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
3114         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
3115         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
3116         * config/i386/t-i386 (i386-d.o): New rule.
3117         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3118         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3119         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
3120         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3121         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
3122         * config/mips/mips-d.c: New file.
3123         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
3124         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
3125         * config/mips/t-mips (mips-d.o): New rule.
3126         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3127         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3128         * config/powerpcspe/powerpcspe-d.c: New file.
3129         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
3130         New prototype.
3131         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
3132         Support GNU D by using 0 as the language type.
3133         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
3134         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
3135         * config/riscv/riscv-d.c: New file.
3136         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
3137         prototype.
3138         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
3139         * config/riscv/t-riscv (riscv-d.o): New rule.
3140         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3141         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3142         * config/rs6000/rs6000-d.c: New file.
3143         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
3144         prototype.
3145         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
3146         Support GNU D by using 0 as the language type.
3147         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
3148         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
3149         * config/s390/s390-d.c: New file.
3150         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
3151         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
3152         * config/s390/t-s390 (s390-d.o): New rule.
3153         * config/sparc/sparc-d.c: New file.
3154         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
3155         prototype.
3156         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
3157         * config/sparc/t-sparc (sparc-d.o): New rule.
3158         * config/t-glibc (glibc-d.o): New rule.
3159         * configure: Regenerated.
3160         * configure.ac (tm_d_file): New variable.
3161         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
3162         * doc/contrib.texi (Contributors): Add self for the D frontend.
3163         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
3164         * doc/install.texi (Configuration): Mention libphobos as an option for
3165         --enable-shared.  Mention d as an option for --enable-languages.
3166         (Testing): Mention check-d as a target.
3167         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
3168         name suffixes.  Mention d as a -x option.
3169         * doc/sourcebuild.texi (Top Level): Mention libphobos.
3170         * doc/standards.texi (Standards): Add section on D language.
3171         * doc/tm.texi: Regenerated.
3172         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
3173         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
3174         * dwarf2out.c (is_dlang): New function.
3175         (gen_compile_unit_die): Use DW_LANG_D for D.
3176         (declare_in_namespace): Return module die for D, instead of adding
3177         extra declarations into the namespace.
3178         (gen_namespace_die): Generate DW_TAG_module for D.
3179         (gen_decl_die): Handle CONST_DECLSs for D.
3180         (dwarf2out_decl): Likewise.
3181         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
3182         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
3183         kinds of aggregates.
3184         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
3185         * genhooks.c: Include d/d-target.def.
3187 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
3189         PR target/85669
3190         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
3191         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
3192         (STACK_DYNAMIC_OFFSET): Likewise.
3194 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
3196         PR target/80024
3197         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
3198         error message.
3200 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
3202         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
3203         return type and other typos.
3205 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
3207         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
3208         a shorter sequence with fewer branches.
3209         (emit_final_str_compare_gpr): Ditto.
3211 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
3213         * config/rs6000/tmmintrin.h: New file.
3214         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
3216 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
3218         * config/rs6000/mmintrin.h: Enable 32bit compilation.
3219         * config/rs6000/xmmintrin.h: Likewise.
3221 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
3223         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
3225 2018-10-26  Richard Biener  <rguenther@suse.de>
3227         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
3228         and wrapper.
3229         (vect_mark_slp_stmts_relevant): Likewise.
3230         (vect_detect_hybrid_slp_stmts): Likewise.
3231         (vect_bb_slp_scalar_cost): Likewise.
3232         (vect_remove_slp_scalar_calls): Likewise.
3234 2018-10-26  Jan Hubicka  <jh@suse.cz>
3236         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
3237         (warn_types_mismatch): Fix walk of DECL_NAME.
3238         (odr_types_equivalent_p): Fix overactive assert.
3240 2018-10-26  Richard Biener  <rguenther@suse.de>
3242         PR tree-optimization/87105
3243         * tree-vectorizer.h (_slp_tree::refcnt): New member.
3244         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
3245         refcnt.
3246         (vect_create_new_slp_node): Initialize refcnt to one.
3247         (bst_traits): Move.
3248         (scalar_stmts_set_t, bst_fail): Remove.
3249         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
3250         (vect_build_slp_tree): Add bst_map argument and lookup
3251         already created SLP nodes.
3252         (vect_print_slp_tree): Handle a SLP graph, print SLP node
3253         addresses.
3254         (vect_slp_rearrange_stmts): Handle a SLP graph.
3255         (vect_analyze_slp_instance): Adjust and free SLP nodes from
3256         the CSE map.  Fix indenting.
3257         (vect_schedule_slp_instance): Add short-cut.
3259 2018-10-26  Martin Liska  <mliska@suse.cz>
3261         PR testsuite/86158
3262         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
3263         addr_expr and not with pointers.
3265 2018-10-26  Jan Hubicka  <jh@suse.cz>
3267         * tree.c (free_lang_data_in_type): Only check main variants.
3268         * ipa-devirt.c (warn_odr): Make static.
3269         (types_same_for_odr): Drop strict variant.
3270         (types_odr_comparable): Likewise.
3271         (odr_or_derived_type_p): Look for main variants.
3272         (odr_name_hasher::equal): Cleanup comment.
3273         (odr_subtypes_equivalent): Add warn and warned arguments; check main
3274         variants.
3275         (type_variants_equivalent_p): break out from ...
3276         (odr_types_equivalent): ... here; go for main variants where needed.
3277         (warn_odr): ... here; turn static.
3278         (warn_types_mismatch): Compare mangled names of main variants.
3279         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
3280         (type_with_linkage_p): Sanity check that we look at main variant.
3281         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
3282         * tree.h (types_same_for_odr): Drop strict argument.
3284 2018-10-26  Richard Biener  <rguenther@suse.de>
3286         PR tree-optimization/87746
3287         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
3288         Simplify and fix WRT strided store groups with size not
3289         equal to step in element count.
3290         (vect_analyze_group_access_1): Dump the whole group.
3292 2018-10-25  Carl Love  <cel@us.ibm.com>
3294         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
3295         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
3296         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
3297         precicion entry for each overloaded builtin.
3298         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
3299         VSCEDPUO): Rename overloaded name.
3300         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
3301         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
3302         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
3303         define_expand for xscmpexqp instruction.
3304         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
3306 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
3307             Jinsong Ji <jji@us.ibm.com>
3309         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
3310         function with vec_sl.
3311         (_mm_slli_epi32): Likewise.
3312         (_mm_slli_epi64): Likewise.
3313         (_mm_srai_epi16): Replace deprecated function with vec_sra.
3314         (_mm_srai_epi32): Likewise.
3315         (_mm_srli_epi16): Replace deprecated function with vec_sr.
3316         (_mm_srli_epi32): Likewise.
3317         (_mm_srli_epi64): Likewise.
3318         (_mm_sll_epi16): Replace deprecated function with vec_sl.
3319         (_mm_sll_epi32): Likewise.
3320         (_mm_sll_epi64): Likewise.
3321         (_mm_sra_epi16): Replace deprecated function with vec_sra.
3322         (_mm_sra_epi32): Likewise.
3323         (_mm_srl_epi16): Replace deprecated function with vec_sr.
3324         (_mm_srl_epi32): Likewise.
3325         (_mm_srl_epi64): Likewise.
3327 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
3328             Jinsong Ji <jji@us.ibm.com>
3330         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
3331         comparison operators with vec_cmp* for compatibility due to
3332         unfortunate history; clean up formatting and use types more
3333         appropriately.
3334         (_mm_sll_epi32): Likewise.
3335         (_mm_sll_epi64): Likewise.
3336         (_mm_srl_epi16): Likewise.
3337         (_mm_srl_epi32): Likewise.
3338         (_mm_srl_epi64): Likewise.
3340 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
3341             Jinsong Ji <jji@us.ibm.com>
3343         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
3344         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
3345         __vector __bool int.  Use vec_cmpgt in preference to deprecated
3346         function vec_vcmpgtfp.
3347         (_mm_max_ps): Likewise.
3349 2018-10-25  Jeff Law  <law@redhat.com>
3351         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
3352         if returning false.
3354 2018-10-25  Martin Sebor  <msebor@redhat.com>
3356         * doc/extend.texi (aligned): Expand attribute description.
3357         (Alignment): Rename section.  Discuss function arguments.
3359 2018-10-25  Jan Hubicka  <jh@suse.cz>
3361         * ipa-devirt.c (main_odr_variant): Remove.
3362         (hash_odr_name, types_same_for_odr, types_odr_comparable,
3363         odr_name_hasher::equal, odr_subtypes_equivalent_p):
3364         Drop use of main_odr_variant.
3365         (add_type_duplicate): Silence confused warnings on integer types.
3366         (get_odr_type): Always look for main variant.
3367         (register_odr_type): Simplify.
3369 2018-10-25  Richard Biener  <rguenther@suse.de>
3371         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
3372         Initialize ng to silence error with release checking bootstrap.
3374 2018-10-25  Richard Biener  <rguenther@suse.de>
3376         * tree-if-conv.c: Include tree-ssa-sccvn.h.
3377         (tree_if_conversion): Run CSE on the if-converted loop body.
3379 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
3381         * config/s390/constraints.md (ZL): New constraint.
3382         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
3383         operands.
3384         * config/s390/s390.md (movdi_larl): Remove.
3385         (movdi_64): Add the LARL alternative.
3387 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
3389         PR bootstrap/87747
3390         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
3391         (RTX_CODE_HWINT_P): New macro.
3392         (rtx_code_size): Use RTX_CODE_HWINT_P ().
3394 2018-10-25  Jan Hubicka  <jh@suse.cz>
3396         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
3397         is anonymous.
3399 2018-10-25  Richard Biener  <rguenther@suse.de>
3401         PR tree-optimization/87665
3402         PR tree-optimization/87745
3403         * tree-vectorizer.h (get_earlier_stmt): Remove.
3404         (get_later_stmt): Pick up UID from the original non-pattern stmt.
3406 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
3408         * options.texi (Deprecated): Move list to Var section.
3410 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
3411             Jinsong Ji <jji@us.ibm.com>
3413         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
3414         __vector long to __vector long long.
3415         (_mm_cvtpd_ps): Likewise.
3416         (_mm_cvttpd_epi32): Likewise.
3417         (_mm_cvtpi32_pd): Likewise.
3418         (_mm_unpackhi_epi64): Likewise.
3419         (_mm_unpacklo_epi64): Likewise.
3421 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
3423         PR rtl-optimization/87720
3424         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
3426 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
3428         * gimple-ssa-isolate-paths.c
3429         (find_implicit_erroneous_behavior): Do not change code if the
3430         pass is running for warnings only.
3431         (find_explicit_erroneous_behavior): Likewise.
3433 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
3435         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
3436         Define as rs6000_mangle_decl_assembler_name.
3437         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
3438         long double to IEEE long double, switch the names of the long
3439         double built-in functions to be <func>f128 instead of <func>l.
3441 2018-10-24  Martin Sebor  <msebor@redhat.com>
3443         * doc/extend.texi (nonnull): List no-argument form.  Reference
3444         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
3446 2018-10-24  Richard Biener  <rguenther@suse.de>
3448         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
3450 2018-10-24  Martin Liska  <mliska@suse.cz>
3452         PR tree-optimization/84436
3453         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
3454         Remove.
3455         (switch_conversion::contains_linear_function_p): New.
3456         (switch_conversion::build_one_array): Support linear
3457         transformation on input.
3458         * tree-switch-conversion.h (struct switch_conversion): Add
3459         contains_linear_function_p declaration.
3461 2018-10-24  Richard Biener  <rguenther@suse.de>
3463         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
3464         if its argument is CONSTANT_CLASS_P.
3466 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
3468         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
3469         it is wrong for forward declarations.
3471 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
3473         * config/s390/s390.c (s390_check_qrst_address): Add the missing
3474         SYMBOL_REF_P () check.
3476 2018-10-24  Richard Biener  <rguenther@suse.de>
3478         PR tree-optimization/87105
3479         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
3480         dump classification.
3481         (vect_analyze_data_ref_accesses): Handle duplicate loads and
3482         stores by splitting the affected group after the fact.
3483         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
3484         fail the SLP build because of size constraints.
3486 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3488         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
3489         * configure: Regenerate.
3490         * config.in: Regenerate.
3491         * varasm.c (mergeable_string_section): Use readonly_data_section
3492         if linker doesn't support SHF_MERGE with alignment > 8.
3493         (mergeable_constant_section): Likewise.
3495 2018-10-24  Richard Biener  <rguenther@suse.de>
3497         PR tree-optimization/84013
3498         * tree-ssa-structalias.c (struct msdi_data): New struct for
3499         marshalling data to walk_stmt_load_store_ops.
3500         (maybe_set_dependence_info): Refactor as callback for
3501         walk_stmt_load_store_ops.
3502         (compute_dependence_clique): Set restrict info on all stmt kinds.
3504 2018-10-24  Martin Liska  <mliska@suse.cz>
3506         * cgraph.c (cgraph_node::dump):
3507         Remove reduntant dumps and make tp_first_run dump more compact.
3509 2018-10-24  Richard Biener  <rguenther@suse.de>
3511         PR tree-optimization/87665
3512         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
3513         to reflect reality.
3515 2018-10-12  Jeff Law  <law@redhat.com>
3517         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
3518         for H8/S.
3520 2018-10-23  Richard Biener  <rguenther@suse.de>
3522         * tree-vrp.c (add_assert_info): Guard dump_printf with
3523         dump_enabled_p.
3524         * gimple-ssa-evrp-analyze.c
3525         (evrp_range_analyzer::record_ranges_from_incoming_edge):
3526         Use value_range::ignore_equivs_equal_p.
3528 2018-10-23  Richard Biener  <rguenther@suse.de>
3530         PR tree-optimization/87105
3531         PR tree-optimization/87608
3532         * passes.def (pass_all_early_optimizations): Add early phi-opt
3533         after dce.
3534         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
3535         addition to debug stmts.
3536         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
3537         and abs replacement early.
3538         * tree-cfg.c (gimple_empty_block_p): Likewise.
3540 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
3542         PR target/86383
3543         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
3544         specified to configure.
3545         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
3547 2018-10-23  Richard Biener  <rguenther@suse.de>
3549         PR tree-optimization/87700
3550         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
3552 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
3554         PR target/87674
3555         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
3556         second argument from __mmask16 to __mmask8.
3557         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
3558         _mm_mask_packs_epi32): Likewise.
3559         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
3560         Likewise.
3561         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
3563 2018-10-23  Richard Biener  <rguenther@suse.de>
3565         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
3567 2018-10-23  Richard Biener  <rguenther@suse.de>
3569         PR tree-optimization/86144
3570         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
3571         over simd attribute.
3573 2018-10-23  Richard Biener  <rguenther@suse.de>
3575         PR tree-optimization/87693
3576         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
3577         the case we do not find the taken edge.
3579 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
3580             Jinsong Ji  <jji@us.ibm.com>
3582         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
3583         (_mm_store_pd): Use unaligned vector type for pointer cast.
3584         (_mm_maskmoveu_si128): Likewise.
3585         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
3586         (_mm_store_ps): Use unaligned vector type for pointer cast.
3588 2018-10-22  Paul Koning  <ni1d@arrl.net>
3590         * symtab.c (symtab_node::increase_alignment): Correct max
3591         alignment check.
3593 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
3595         PR tree-optimization/87633
3596         * match.pd: Do not generate unordered integer comparisons.
3598 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
3600         PR rtl-optimization/87600
3601         * combine.c: Add include of expr.h.
3602         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
3603         register to a pseudo.
3604         (make_more_copies): New function, add a copy to a new pseudo after
3605         the moves from hard registers into pseudos.
3606         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
3607         later.  Call make_more_copies.
3609 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
3611         * lra-constraints.c (process_alt_operands): New local array,
3612         matching_early_clobber.  Check matching_early_clobber before
3613         decrementing reject, and set matching_early_clobber after.
3615 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
3617         PR target/87598
3618         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
3619         call output_addr_const and hope for the best.
3621 2018-10-22  Richard Biener  <rguenther@suse.de>
3623         * gimple-ssa-evrp-analyze.c
3624         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
3625         smarter about what ranges to use.
3626         * tree-vrp.c (add_assert_info): Dump here.
3627         (register_edge_assert_for_2): Instead of here at multiple but
3628         not all places.
3630         * gcc.dg/tree-ssa/evrp12.c: New testcase.
3631         * gcc.dg/predict-6.c: Adjust.
3632         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
3633         * gcc.dg/tree-ssa/vrp02.c: Likewise.
3634         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
3636 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
3637         Richard Biener  <rguenther@suse.de>
3639         * bitmap.h: Update data structure documentation, including a
3640         description of bitmap views as either linked-lists or splay trees.
3641         (struct bitmap_element_def): Update comments for splay tree bitmaps.
3642         (struct bitmap_head_def): Likewise.
3643         (bitmap_list_view, bitmap_tree_view): New prototypes.
3644         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
3645         tree_form fields.
3646         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
3647         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
3648         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
3649         released bitmap element here.
3650         (bitmap_element_free): Remove.
3651         (bitmap_elt_clear_from): Work on splay tree bitmaps.
3652         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
3653         this function similar ones such that linked-list bitmap implementation
3654         functions are grouped.
3655         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
3656         and moved for grouping.
3657         (bitmap_list_insert_element_after): Renamed from
3658         bitmap_elt_insert_after, and moved for grouping.
3659         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
3660         (bitmap_tree_link_left, bitmap_tree_link_right,
3661         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
3662         bitmap_tree_link_element, bitmap_tree_unlink_element,
3663         bitmap_tree_find_element): New functions for splay-tree bitmap
3664         implementation.
3665         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
3666         Renamed and moved, see above entries.
3667         (bitmap_tree_listify_from): New function to convert part of a splay
3668         tree bitmap to a linked-list bitmap.
3669         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
3670         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
3671         (bitmap_find_bit): Remove.
3672         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
3673         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
3674         Handle splay tree bitmaps.
3675         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
3676         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
3677         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
3678         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
3679         bitmap_intersect_compl_p, bitmap_ior_and_compl,
3680         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
3681         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
3682         corresponding changes to use linked-list specific bitmap_element
3683         manipulation functions as applicable for efficiency.
3684         (bitmap_tree_to_vec): New function.
3685         (debug_bitmap_elt_file): New function split out from ...
3686         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
3687         (bitmap_print): Likewise.
3689         PR tree-optimization/63155
3690         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
3691         SSA edge worklists.
3692         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
3693         in tree-view.
3695 2018-10-22  Martin Liska  <mliska@suse.cz>
3697         PR tree-optimization/87686
3698         Revert
3699         2018-08-29  Martin Liska  <mliska@suse.cz>
3701         * tree-switch-conversion.c (switch_conversion::expand):
3702         Strenghten assumption about gswitch statements.
3704 2018-10-22  Martin Liska  <mliska@suse.cz>
3706         * ipa-icf.c (sem_item::compare_attributes): Remove.
3707         (sem_item::compare_referenced_symbol_properties): Use
3708         attribute_list_equal instead.
3709         (sem_function::equals_wpa): Likewise.
3710         * ipa-icf.h: Remove compare_attributes.
3712 2018-10-22  Richard Biener  <rguenther@suse.de>
3714         PR middle-end/87682
3715         * mem-stats.h (mem_usage::operator==): Fix pasto.
3717 2018-10-22  Richard Biener  <rguenther@suse.de>
3719         PR tree-optimization/87640
3720         * tree-vrp.c (set_value_range_with_overflow): Decompose
3721         incomplete result.
3722         (extract_range_from_binary_expr_1): Adjust.
3724 2018-10-22  Martin Jambor  <mjambor@suse.cz>
3726         * tree-eh.h (stmt_could_throw_p): Add function parameter.
3727         (stmt_can_throw_external): Likewise.
3728         (stmt_can_throw_internal): Likewise.
3729         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
3730         (lower_eh_constructs_2): Likewise.
3731         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
3732         (stmt_can_throw_external): Likewise.
3733         (stmt_can_throw_internal): Likewise.
3734         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
3735         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
3736         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
3737         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
3738         (pass_lower_eh_dispatch::execute): Pass cfun to
3739         stmt_can_throw_external.
3740         (cleanup_empty_eh): Likewise.
3741         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
3742         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
3743         stmt_can_throw_external instead of pushing it to cfun.
3744         (symbol_table::create_edge): Likewise.
3745         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
3746         stmt_can_throw_internal.
3747         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
3748         to stmt_could_throw_p.
3749         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
3750         stmt_can_throw_internal.
3751         (pass_store_merging::execute): Likewise.
3752         * gimple-ssa-strength-reduction.c
3753         (find_candidates_dom_walker::before_dom_children): Pass cfun to
3754         stmt_could_throw_p.
3755         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
3756         stmt_can_throw_internal.
3757         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
3758         to stmt_can_throw_external.
3759         (check_stmt): Pass cfun to stmt_could_throw_p.
3760         (check_stmt): Pass cfun to stmt_can_throw_external.
3761         (pass_nothrow::execute): Likewise.
3762         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
3763         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
3764         stmt_can_throw_internal.
3765         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
3766         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
3767         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
3768         * tree-complex.c (expand_complex_libcall): Pass cfun to
3769         stmt_could_throw_p and to stmt_can_throw_internal.
3770         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
3771         * tree-inline.c (copy_edges_for_bb): Likewise.
3772         (maybe_move_debug_stmts_to_successors): Likewise.
3773         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
3774         stmt_could_throw_p.
3775         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
3776         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
3777         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
3778         stmt_can_throw_internal.
3779         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
3780         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
3781         stmt_could_throw_p.
3782         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
3783         stmt_can_throw_internal.
3784         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
3785         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
3786         stmt_could_throw_p.
3787         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
3788         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
3789         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
3790         (convert_mult_to_fma_1): Likewise.
3791         (convert_to_divmod): Likewise.
3792         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
3793         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
3794         * tree-ssa-propagate.c
3795         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
3796         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
3797         (maybe_optimize_range_tests): Likewise.
3798         (linearize_expr_tree): Likewise.
3799         (reassociate_bb): Likewise.
3800         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3801         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
3802         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
3803         (handle_char_store): Likewise.
3804         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
3805         stmt_can_throw_internal.
3806         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
3807         stmt_could_throw_p.
3808         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
3809         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
3810         (vectorizable_simd_clone_call): Likewise.
3811         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
3812         (gimple_stringop_fixed_value): Likewise.
3814 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
3816         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
3817         literal pool references.
3818         (s390_check_qrst_address): Adapt to the new behavior of
3819         s390_loadrelative_operand_p ().
3821 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3823         PR target/72782
3824         * config/i386/sse.md (*andnot<mode>3_bcst): New.
3826 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3828         PR target/72782
3829         * config/i386/sse.md (*<code><mode>3_bcst): New.
3831 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3833         PR target/72782
3834         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
3835         V4DI, V16SI and V8DI.
3836         (*sub<mode>3<mask_name>_bcst): New.
3837         (*add<mode>3<mask_name>_bcst): Likewise.
3839 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
3840             Jinsong Ji  <jji@us.ibm.com>
3842         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
3843         __m64 with __vector unsigned long long for compatibility.
3844         (_mm_movemask_epi8): Likewise.
3845         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
3846         (_mm_cvttps_pi32): Likewise.
3847         (_mm_cvtpi32_ps): Likewise.
3848         (_mm_cvtps_pi16): Likewise.
3849         (_mm_loadh_pi): Likewise.
3850         (_mm_storeh_pi): Likewise.
3851         (_mm_movehl_ps): Likewise.
3852         (_mm_movelh_ps): Likewise.
3853         (_mm_loadl_pi): Likewise.
3854         (_mm_storel_pi): Likewise.
3855         (_mm_movemask_ps): Likewise.
3856         (_mm_shuffle_pi16): Likewise.
3858 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3860         PR target/72782
3861         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
3862         __builtin_ia32_vfnmsubpd512_mask.
3863         (_mm512_mask_fnmsub_round_pd): Likewise.
3864         (_mm512_fnmsub_pd): Likewise.
3865         (_mm512_mask_fnmsub_pd): Likewise.
3866         (_mm512_maskz_fnmsub_round_pd): Use
3867         __builtin_ia32_vfnmsubpd512_maskz.
3868         (_mm512_maskz_fnmsub_pd): Likewise.
3869         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
3870         (_mm512_mask_fnmsub_round_ps): Likewise.
3871         (_mm512_fnmsub_ps): Likewise.
3872         (_mm512_mask_fnmsub_ps): Likewise.
3873         (_mm512_maskz_fnmsub_round_ps): Use
3874         __builtin_ia32_vfnmsubps512_maskz.
3875         (_mm512_maskz_fnmsub_ps): Likewise.
3876         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
3877         __builtin_ia32_vfnmsubpd256_mask.
3878         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
3879         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
3880         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
3881         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
3882         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
3883         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
3884         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
3885         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
3886         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
3887         __builtin_ia32_vfnmsubpd.
3888         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
3889         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
3890         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
3891         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
3892         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
3893         * config/i386/i386-builtin.def: Add
3894         __builtin_ia32_vfnmsubpd256_mask,
3895         __builtin_ia32_vfnmsubpd256_maskz,
3896         __builtin_ia32_vfnmsubpd128_mask,
3897         __builtin_ia32_vfnmsubpd128_maskz,
3898         __builtin_ia32_vfnmsubps256_mask,
3899         __builtin_ia32_vfnmsubps256_maskz,
3900         __builtin_ia32_vfnmsubps128_mask,
3901         __builtin_ia32_vfnmsubps128_maskz,
3902         __builtin_ia32_vfnmsubpd512_mask,
3903         __builtin_ia32_vfnmsubpd512_maskz,
3904         __builtin_ia32_vfnmsubps512_mask,
3905         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
3906         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
3907         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
3908         __builtin_ia32_vfnmsubpd256.
3909         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
3910         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
3911         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
3912         Likewise.
3913         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
3914         Likewise.
3915         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
3916         Likewise.
3917         (fmai_vmfnmsub_<mode><round_name>): Likewise.
3919 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3921         PR target/72782
3922         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
3923         __builtin_ia32_vfnmaddpd512_mask.
3924         (_mm512_mask_fnmadd_round_pd): Likewise.
3925         (_mm512_fnmadd_pd): Likewise.
3926         (_mm512_mask_fnmadd_pd): Likewise.
3927         (_mm512_maskz_fnmadd_round_pd): Use
3928         __builtin_ia32_vfnmaddpd512_maskz.
3929         (_mm512_maskz_fnmadd_pd): Likewise.
3930         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
3931         (_mm512_mask_fnmadd_round_ps): Likewise.
3932         (_mm512_fnmadd_ps): Likewise.
3933         (_mm512_mask_fnmadd_ps): Likewise.
3934         (_mm512_maskz_fnmadd_round_ps): Use
3935         __builtin_ia32_vfnmaddps512_maskz.
3936         (_mm512_maskz_fnmadd_ps): Likewise.
3937         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
3938         __builtin_ia32_vfnmaddpd256_mask.
3939         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
3940         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
3941         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
3942         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
3943         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
3944         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
3945         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
3946         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
3947         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
3948         __builtin_ia32_vfnmaddpd.
3949         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
3950         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
3951         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
3952         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
3953         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
3954         * config/i386/i386-builtin.def: Add
3955         __builtin_ia32_vfnmaddpd256_mask,
3956         __builtin_ia32_vfnmaddpd256_maskz,
3957         __builtin_ia32_vfnmaddpd128_mask,
3958         __builtin_ia32_vfnmaddpd128_maskz,
3959         __builtin_ia32_vfnmaddps256_mask,
3960         __builtin_ia32_vfnmaddps256_maskz,
3961         __builtin_ia32_vfnmaddps128_mask,
3962         __builtin_ia32_vfnmaddps128_maskz,
3963         __builtin_ia32_vfnmaddpd512_mask,
3964         __builtin_ia32_vfnmaddpd512_maskz,
3965         __builtin_ia32_vfnmaddps512_mask,
3966         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
3967         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
3968         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
3969         __builtin_ia32_vfnmaddpd256.
3970         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
3971         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
3972         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
3973         Likewise.
3974         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
3975         Likewise.
3976         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
3977         Likewise.
3978         (fmai_vmfnmadd_<mode><round_name>): Likewise.
3980 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3982         PR target/72782
3983         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
3984         __builtin_ia32_vfmsubpd512_mask.
3985         (_mm512_mask_fmsub_round_pd): Likewise.
3986         (_mm512_fmsub_pd): Likewise.
3987         (_mm512_mask_fmsub_pd): Likewise.
3988         (_mm512_maskz_fmsub_round_pd): Use
3989         __builtin_ia32_vfmsubpd512_maskz.
3990         (_mm512_maskz_fmsub_pd): Likewise.
3991         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
3992         (_mm512_mask_fmsub_round_ps): Likewise.
3993         (_mm512_fmsub_ps): Likewise.
3994         (_mm512_mask_fmsub_ps): Likewise.
3995         (_mm512_maskz_fmsub_round_ps): Use
3996         __builtin_ia32_vfmsubps512_maskz.
3997         (_mm512_maskz_fmsub_ps): Likewise.
3998         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
3999         __builtin_ia32_vfmsubpd256_mask.
4000         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
4001         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
4002         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
4003         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
4004         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
4005         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
4006         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
4007         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
4008         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
4009         __builtin_ia32_vfmsubpd.
4010         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
4011         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
4012         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
4013         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
4014         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
4015         * config/i386/i386-builtin.def: Add
4016         __builtin_ia32_vfmsubpd256_mask,
4017         __builtin_ia32_vfmsubpd256_maskz,
4018         __builtin_ia32_vfmsubpd128_mask,
4019         __builtin_ia32_vfmsubpd128_maskz,
4020         __builtin_ia32_vfmsubps256_mask,
4021         __builtin_ia32_vfmsubps256_maskz,
4022         __builtin_ia32_vfmsubps128_mask,
4023         __builtin_ia32_vfmsubps128_maskz,
4024         __builtin_ia32_vfmsubpd512_mask,
4025         __builtin_ia32_vfmsubpd512_maskz,
4026         __builtin_ia32_vfmsubps512_mask,
4027         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
4028         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
4029         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
4030         __builtin_ia32_vfmsubpd256.
4031         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
4032         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
4033         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
4034         Likewise.
4035         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
4036         Likewise.
4037         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
4038         Likewise.
4039         (fmai_vmfmsub_<mode><round_name>): Likewise.
4041 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
4043         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
4044         Remove plus.  Renamed to ...
4045         (*sub<mode>3<mask_name>_bcst): This.
4046         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
4047         (*add<mode>3<mask_name>_bcst): This.
4049 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
4051         PR target/72782
4052         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
4054 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
4056         PR target/87662
4057         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
4058         (_mm_or_epi32): Likewise.
4059         (_mm256_xor_epi32): Likewise.
4060         (_mm_xor_epi32): Likewise.
4061         (_mm256_or_epi64): Likewise.
4062         (_mm_or_epi64): Likewise.
4063         (_mm256_xor_epi64): Likewise.
4064         (_mm_xor_epi64): Likewise.
4066 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
4068         PR target/72782
4069         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
4071 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
4073         PR middle-end/87647
4074         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
4076 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
4078         * doc/ux.texi: Move @section directly after @node.
4080 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
4082         PR middle-end/85488
4083         PR middle-end/87649
4084         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
4085         depend closely nested inside of loop with ordered clause with
4086         a parameter.
4088 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
4090         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
4091         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
4092         * doc/ux.texi: New file.
4094 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
4096         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
4097         be the first CR field allocated.
4099 2018-10-19  Richard Biener  <rguenther@suse.de>
4101         PR target/87657
4102         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
4103         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
4105 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
4107         PR target/72782
4108         * config/i386/sse.md
4109         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
4110         (*add<mode>3<mask_name>_bcst_2): Likewise.
4112 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
4114         * config/i386/sse.md
4115         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
4116         Replace nonimmediate_operand with register_operand.
4117         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
4118         Likewise.
4119         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
4120         Likewise.
4122 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
4124         PR rtl-optimization/87596
4125         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
4126         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
4127         for instructions in FROM..TO range.
4129 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
4131         * cfgexpand.c (expand_one_var): Use specific wording in error message
4132         for non-local frame variables.
4133         * stor-layout.c (layout_decl): Do not issue a warning for them.
4135 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
4137         * haifa-sched.c (priority): Add force_recompute parameter.
4138         (apply_replacement): Call priority () with force_recompute = true.
4139         (restore_pattern): Likewise.
4141 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4143         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
4144         HOST_BITS_PER_WIDE_INT.
4145         (test_vector_ops_duplicate): Likewise.
4147 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4149         PR target/72782
4150         * config/i386/sse.md (VF_AVX512): New.
4151         (avx512bcst): Likewise.
4152         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
4153         Likewise.
4154         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
4155         Likewise.
4156         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
4157         Likewise.
4159 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
4161         * doc/invoke.texi (-dumpversion): Improve grammar.
4162         (-dumpfullversion): Make more consistent with -dumpversion.
4164 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
4166         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
4167         Set JUMP_LABEL to the jump insn.
4168         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
4169         Predict emitted jump and add label to jump insn.
4171 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
4173         PR tree-optimization/87562
4174         * input.c (get_substring_ranges_for_loc): Use
4175         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
4176         getting the linemap for the endpoint.  Verify that it's either
4177         in the same linemap as the start point's spelling location, or
4178         at least in the same file.
4180 2018-10-18  Richard Biener  <rguenther@suse.de>
4182         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
4183         feed width-specific load/store costs through ix86_vec_cost.
4184         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
4185         (k8_cost): Likewise.
4186         (bdver_cost): Likewise.
4187         (znver1_cost): Likewise.
4188         (btver1_cost): Likewise.
4189         (btver2_cost): Likewise.
4191 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4193         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
4194         to simplify subreg of vec_merge.
4196 2018-10-18  Richard Biener  <rguenther@suse.de>
4198         * config/i386/i386.c: Fix costing of vector FMA.
4200 2018-10-18  Richard Biener  <rguenther@suse.de>
4202         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
4203         and argument.
4204         (ix86_builtin_vectorization_cost): For vec_construct properly
4205         cost insertion into SSE regs.
4206         (...): Adjust calls to ix86_vec_cost.
4208 2018-10-18  Richard Biener  <rguenther@suse.de>
4210         PR middle-end/87087
4211         Revert
4212         2018-02-07  Richard Biener  <rguenther@suse.de>
4214         PR tree-optimization/84204
4215         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4216         this place.
4218 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4220         PR target/87537
4221         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
4222         of vec_duplicate.
4223         (test_vector_ops_duplicate): Add test for a scalar subreg of a
4224         VEC_MERGE of a VEC_DUPLICATE.
4226 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
4228         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
4229         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
4230         * doc/standards.texi (C Language): Document C2X.
4231         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
4232         (rl78_option_override): Handle "GNU C2X" language name.
4234 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
4236         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
4237         Document C17 as published in 2018.
4239 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
4241         PR middle-end/87623
4242         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
4243         bail out if both sides do not have the same storage order.
4245 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
4247         * bitmap.c (bitmap_head::dump): New.
4248         * bitmap.h (bitmap_head): Add dump().
4249         * gimple-ssa-evrp-analyze.c
4250         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
4251         (evrp_range_analyzer::set_ssa_range_info): Same.
4252         (evrp_range_analyzer::record_ranges_from_phis): Same.
4253         (evrp_range_analyzer::record_ranges_from_stmt): Same.
4254         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
4255         * gimple-ssa-sprintf.c (get_int_range): Same.
4256         (format_integer): Same.
4257         (sprintf_dom_walker::handle_gimple_call): Same.
4258         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
4259         (ipcp_vr_lattice::top_p): Same.
4260         (ipcp_vr_lattice::bottom_p): Same.
4261         (ipcp_vr_lattice::set_to_bottom): Same.
4262         (ipa_vr_operation_and_type_effects): Same.
4263         (propagate_vr_across_jump_function): Same.
4264         (ipcp_store_vr_results): Same.
4265         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
4266         (ipa_print_node_jump_functions_for_edge): Same.
4267         (ipa_get_value_range): Same.
4268         (ipa_compute_jump_functions_for_edge): Same.
4269         (ipa_write_jump_function): Same.
4270         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
4271         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
4272         Same.
4273         * vr-values.c (set_value_range_to_nonnegative): Same.
4274         (set_value_range_to_truthvalue): Same.
4275         (vr_values::get_value_range): Same.
4276         (vr_values::set_defs_to_varying): Same.
4277         (vr_values::update_value_range): Same.
4278         (symbolic_range_based_on_p): Same.
4279         (vr_values::op_with_boolean_value_range_p): Same.
4280         (vr_values::extract_range_for_var_from_comparison_expr): Same.
4281         (vr_values::extract_range_from_ssa_name): Same.
4282         (vr_values::extract_range_from_binary_expr): Same.
4283         (vr_values::extract_range_from_unary_expr): Same.
4284         (vr_values::extract_range_from_cond_expr): Same.
4285         (vr_values::extract_range_from_comparison): Same.
4286         (vr_values::check_for_binary_op_overflow): Same.
4287         (vr_values::extract_range_basic): Same.
4288         (vr_values::extract_range_from_assignment): Same.
4289         (compare_ranges): Same.
4290         (compare_range_with_value): Same.
4291         (vr_values::adjust_range_with_scev): Same.
4292         (vrp_valueize): Same.
4293         (vrp_valueize_1): Same.
4294         (vr_values::get_vr_for_comparison): Same.
4295         (vr_values::compare_name_with_value): Same.
4296         (vr_values::compare_names): Same.
4297         (vr_values::vrp_evaluate_conditional): Same.
4298         (find_case_label_ranges): Same.
4299         (vr_values::vrp_visit_switch_stmt): Same.
4300         (vr_values::extract_range_from_phi_node): Same.
4301         (vr_values::simplify_div_or_mod_using_ranges): Same.
4302         (vr_values::simplify_bit_ops_using_ranges): Same.
4303         (test_for_singularity): Same.
4304         (range_fits_type_p): Same.
4305         (vr_values::simplify_cond_using_ranges_1): Same.
4306         (vr_values::simplify_switch_using_ranges): Same.
4307         (vr_values::simplify_float_conversion_using_ranges): Same.
4308         (vr_values::two_valued_val_range_p): Same.
4309         (vr_values::add_equivalence): Move to value_range::equiv_add.
4310         * vr-values.h (vr_values::add_equivalence): Remove.
4311         (VR_INITIALIZER): Remove.
4312         * tree-vrp.c (value_range::set): New.
4313         (value_range::equiv_add): New.
4314         (value_range::value_range): New.
4315         (value_range::deep_copy): New.
4316         (value_range::check): New.
4317         (value_range::equal_p): New.
4318         (value_range::ignore_equivs_equal_p): New.
4319         (value_range::operator==): New.
4320         (value_range::operator!=): New.
4321         (value_range::symbolic_p): New.
4322         (value_range::numeric_p): New.
4323         (value_range::set_undefined): New.
4324         (value_range::set_varying): New.
4325         (value_range::may_contain_p): New.
4326         (value_range::equiv_clear): New.
4327         (value_range::singleton_p): New.
4328         (value_range::intersect): New.
4329         (value_range::dump): New.
4330         (value_range::set_and_canonicalize): New.
4331         (set_value_range): Adjust for value_range API.
4332         (set_value_range_to_undefined): Same.
4333         (set_value_range_to_varying): Same.
4334         (set_and_canonicalize_value_range): Same.
4335         (set_value_range_to_nonnull): Same.
4336         (set_value_range_to_null): Same.
4337         (range_is_null): Same.
4338         (range_is_nonnull): Same.
4339         (range_int_cst_p): Same.
4340         (range_int_cst_singleton_p): Same.
4341         (symbolic_range_p): Same.
4342         (range_includes_zero_p): Same.
4343         (value_range_constant_singleton): Same.
4344         (vrp_set_zero_nonzero_bits): Same.
4345         (ranges_from_anti_range): Same.
4346         (extract_range_into_wide_ints): Same.
4347         (extract_range_from_multiplicative_op): Same.
4348         (set_value_range_with_overflow): Same.
4349         (extract_range_from_binary_expr_1): Same.
4350         (extract_range_from_unary_expr): Same.
4351         (dump_value_range): Same.
4352         (debug_value_range): Same.
4353         (vrp_prop::check_array_ref): Same.
4354         (vrp_prop::check_mem_ref): Same.
4355         (vrp_prop::vrp_initialize): Same.
4356         (vrp_prop::visit_stmt): Same.
4357         (intersect_ranges): Same.
4358         (vrp_prop::visit_phi): Same.
4359         (vrp_prop::vrp_finalize): Same.
4360         (determine_value_range_1): Same.
4361         (determine_value_range): Same.
4362         (vrp_intersect_ranges_1): Rename to...
4363         (vrp_intersect_1): this.
4364         (vrp_intersect_ranges): Rename to...
4365         (value_range::intersect_helper): ...this.
4366         (vrp_meet_1): Rename to...
4367         (value_range::union_helper): ...this.
4368         (vrp_meet): Rename to...
4369         (value_range::union_): ...this.
4370         (copy_value_range): Remove.
4371         * tree-vrp.h (struct value_range): Rewrite into a proper class.
4372         (value_range::vrtype): New.
4373         (value_range::type): New.
4374         (value_range::equiv): New.
4375         (value_range::min): New.
4376         (value_range::max): New.
4377         (value_range::varying_p): New.
4378         (value_range::undefined_p): New.
4379         (value_range::null_p): New.
4380         (value_range::equiv_add): New.
4381         (copy_value_range): Remove.
4383 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
4385         * Makefile.in (SELFTEST_TARGETS): New.
4386         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
4387         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
4388         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
4389         c/Make-lang.in.
4390         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
4391         (selftest-c++-gdb, selftest-c++-valgrind): Move to
4392         cp/Make-lang.in.
4393         * configure: Regenerate.
4394         * configure.ac (selftest_languages): New.
4396 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
4398         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
4399         overflow wraps argument.
4400         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
4401         wide_int_range_multiplicative_op.
4402         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
4403         overflow wraps argument.
4404         (wide_int_range_multiplicative_op): Same.
4405         (wide_int_range_lshift): Same.
4406         (wide_int_range_div): Same.
4407         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
4408         (wide_int_range_lshift): Same.
4409         (wide_int_range_div): Same.
4411 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
4413         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
4414         use sign as argument.
4415         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
4416         wide_int_range_shift_undefined_p.
4418 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4420         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
4421         Rename to...
4422         (@despeculate_copy<ALLI_TI:mode>): ... This.
4423         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
4424         switch statement.
4426 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4428         * config.gcc: Obsolete *-*-solaris2.10*.
4429         * doc/install.texi (Specific, *-*-solaris2*): Document it.
4431 2018-10-12  Jeff Law  <law@redhat.com>
4433         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
4434         reg + sym +- const_int addressing modes.
4436 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
4438         * common.opt (fdiagnostics-minimum-margin-width=): New option.
4439         * diagnostic-show-locus.c (layout::layout): Apply the minimum
4440         margin width.
4441         (layout::start_annotation_line): Only print up to 3 of the
4442         margin character, to avoid touching the left-hand side.
4443         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
4444         minimum margin width, as set by test_diagnostic_context's ctor.
4445         (selftest::test_fixit_insert_containing_newline): Likewise.
4446         (selftest::test_fixit_insert_containing_newline_2): Likewise.
4447         (selftest::test_line_numbers_multiline_range): Clear
4448         dc.min_margin_width.
4449         * diagnostic.c (diagnostic_initialize): Initialize
4450         min_margin_width.
4451         * diagnostic.h (struct diagnostic_context): Add field
4452         "min_margin_width".
4453         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
4454         * opts.c (common_handle_option): Handle
4455         OPT_fdiagnostics_minimum_margin_width_.
4456         * selftest-diagnostic.c
4457         (selftest::test_diagnostic_context::test_diagnostic_context):
4458         Initialize min_margin_width to 6.
4459         * toplev.c (general_init): Initialize global_dc->min_margin_width.
4461 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
4463         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
4464         Fix usage of "error_at_rich_loc" in the comment.
4466 2018-10-15  Renlin Li  <renlin.li@arm.com>
4468         PR target/87563
4469         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
4470         if-conversioned loop when it contains ifn with types not
4471         supported by backend.
4472         * internal-fn.c (expand_direct_optab_fn): Add an assert.
4473         (direct_internal_fn_supported_p): New helper function.
4474         * internal-fn.h (direct_internal_fn_supported_p): Declare.
4476 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
4478         PR target/87572
4479         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
4480         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
4481         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
4483 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
4485         PR tree-optimization/87022
4486         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
4487         bits in dist vector rather than the first one.
4489 2018-10-15  Richard Biener  <rguenther@suse.de>
4491         PR middle-end/87610
4492         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
4493         (visit_loadstore): When a used restrict tag escaped verify that
4494         the points-to solution of "other" pointers do not include
4495         escaped.
4496         (compute_dependence_clique): If a used restrict tag escaped
4497         communicated that down to visit_loadstore.
4499 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
4501         * config/s390/s390.c (s390_expand_vec_init): Force vector element
4502         into reg if it isn't a general operand.
4504 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
4506         PR target/87599
4507         * config/i386/sse.md (*vec_dupv2di): Add register source to
4508         movddup.
4510 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
4512         PR target/87572
4513         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
4514         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
4515         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
4516         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
4518 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
4520         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
4521         (notice_args_size): Set it in the current trace if no insn that can
4522         throw internally has been seen yet.
4523         (connect_traces): When connecting args_size between traces, allow the
4524         incoming values not to match if there is an insn setting it before the
4525         first insn that can throw internally; in that case, force the creation
4526         of a CFI note on this latter insn.
4528 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
4530         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
4531         the base class, because using the injected-class-name was not clearly
4532         specified until DR 176.
4534 2018-10-12  Paul Koning  <ni1d@arrl.net>
4536         * config/pdp11/pdp11.md (doloop_end): New expander.
4537         (doloop_end_insn): renamed from "doloop_end".
4538         (addqi3): New pattern.
4539         (subqi3): New pattern.
4540         * config/pdp11/predicates.md (incdec_operand): New predicate.
4542 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
4544         PR middle-end/81376
4545         * real.c (format_helper::can_represent_integral_type_p): New function
4546         * real.h (format_helper::can_represent_integral_type_p): Ditto.
4547         * match.pd: New pattern.
4549 2018-10-12  Alexandre Oliva <oliva@adacore.com>
4551         * configure.ac: Introduce --enable-large-address-aware
4552         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
4553         * doc/install.texi: Document it.
4554         * configure, config.in: Rebuilt.
4555         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
4556         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
4557         (LINK_SPEC): Insert it.
4558         * config/i386/mingw-w64.h: Likewise.
4560         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
4562 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
4564         PR rtl-optimization/87600
4565         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
4567 2018-10-12  Paul Koning  <ni1d@arrl.net>
4569         * doc/md.texi (doloop_end): Document that the pattern code may
4570         need to check operand mode.
4572 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
4574         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
4575         to zero-extend between int and floating-point registers.
4576         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
4577         ldp into floating-point registers.  Add type and arch attributes.
4578         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
4579         Use f_loads for type attribute.
4581 2018-10-11  Martin Sebor  <msebor@redhat.com>
4583         * doc/extend.texi (attribute packed): Correct typos.
4585 2018-10-11  Martin Sebor  <msebor@redhat.com>
4587         * doc/extend.texi (attribute flatten): Mention interaction with
4588         noinline.
4590 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
4592         PR target/87156
4593         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4594         Set new_decl virtual flag to zero.
4596 2018-10-11  Martin Sebor  <msebor@redhat.com>
4598         PR middle-end/87593
4599         * doc/extend.texi (attribute format_arg): Discuss using multiple
4600         attributes on a single function.
4602 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
4604         PR tree-optimization/86829
4605         * match.pd (sin (atan (x))): New simplification rules.
4606         (cos (atan (x))): Likewise.
4607         * real.c (build_sinatan_real): New function.
4608         * real.h (build_sinatan_real): Prototype.
4610 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
4612         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
4613         function.
4614         (fold_mergeeo_helper): New helper function.
4615         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
4616         intrinsics.  Correct some whitespace indentation issues.
4618 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
4620         PR target/87511
4621         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
4622         Use HOST_WIDE_INT_1U for shift.
4624 2018-10-11  Doug Rupp  <rupp@adacore.com>
4625             Olivier Hainque  <hainque@adacore.com>
4627         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
4628         Pass --relax to the linker for RTPs.
4629         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
4631 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
4632             Jan Hubicka  <jh@suse.cz>
4633             Martin Jambor  <mjambor@suse.cz>
4635         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
4636         the same elements are repeated rather than printing all of them.
4637         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
4638         "repeated" elements.
4639         * read-rtl-function.c (test_loading_repeat): New function.
4640         (read_rtl_function_c_tests): Call test_loading_repeat.
4641         * rtl-tests.c (test_dumping_repeat): New function.
4642         (rtl_tests_c_tests): Call test_dumping_repeat.
4644 2018-10-11  Richard Biener  <rguenther@suse.de>
4646         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
4647         bdver?_cost): Unify to ...
4648         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
4649         * config/i386/i386.c (processor_cost_table): Adjust.
4651 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
4653         PR middle-end/87574
4654         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
4655         the thunk when expanding to GIMPLE.
4657 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4659         * varasm.c (mergeable_string_section): Don't try to move zero-length
4660         strings to the merge section.
4662 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
4664         PR target/87573
4665         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
4667 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
4669         PR target/87550
4670         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
4671         to special_args set.
4673 2018-10-10  Richard Biener  <rguenther@suse.de>
4675         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
4676         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
4677         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
4678         and recursing and pattern terminating the recursion on SSE
4679         vector width using ix86_expand_reduc.
4680         (reduc_sminmax_scal_<mode>): Split into part reducing to half
4681         width and recursing and SSE2 vector variant doing the final
4682         reduction with ix86_expand_reduc.
4683         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
4684         with terminating the recursion at AVX level, splitting that
4685         to SSE there.
4687 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
4689         * genmatch.c (error_cb): Rename to...
4690         (diagnostic_cb): ...this, converting int params to enums.
4691         (fatal_at): Update for renaming.
4692         (warning_at): Likewise.
4693         (main): Likewise.
4694         * input.c (selftest::ebcdic_execution_charset::apply):
4695         Update for renaming of...
4696         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
4697         to...
4698         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
4699         converting level and reason to enums.
4700         (class selftest::lexer_error_sink): Rename to...
4701         (class selftest::lexer_test_options): ...this, renaming field
4702         "m_errors" to "m_diagnostics".
4703         (selftest::lexer_test_options::apply): Update for renaming of...
4704         (selftest::lexer_test_options::on_error): ...this, renaming to...
4705         (selftest::lexer_test_options::on_diagnostic): ...this
4706         converting level and reason to enums.
4707         (selftest::test_lexer_string_locations_raw_string_unterminated):
4708         Update for renamings.
4709         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
4710         "reason".
4712 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
4714         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
4715         * config/rs6000/pmmintrin.h: New file.
4717 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
4719         PR tree-optimization/86659
4720         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
4722 2018-10-09  Richard Biener  <rguenther@suse.de>
4724         PR tree-optimization/63155
4725         * tree-ssa-structalias.c: Include tree-ssa.h.
4726         (get_constraint_for_ssa_var): For undefs return nothing_id.
4727         (find_func_aliases): Cleanup PHI handling.
4729 2018-10-09  Richard Biener  <rguenther@suse.de>
4731         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
4732         replacements.
4734 2018-10-09  Martin Liska  <mliska@suse.cz>
4736         * asan.c (asan_emit_stack_protection): If a stack variable
4737         is located in a same file as current function, then emit
4738         line info into variable definition string.
4740 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4742         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
4743         information.
4745 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4747         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
4748         on the thunk.
4750 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4752         PR tree-optimization/86659
4753         * gimple-match.h (struct gimple_match_op): Add reverse field.
4754         (gimple_match_op::set_op): New overloaded method.
4755         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
4756         the REF_REVERSE_STORAGE_ORDER flag on the value.
4757         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
4758         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
4759         
4760 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
4762         PR middle-end/63155
4763         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
4764         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
4766 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
4768         PR target/87517
4769         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
4770         Defined with __builtin_ia32_vfmaddsubpd512_mask.
4772 2018-10-08  Richard Biener  <rguenther@suse.de>
4774         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
4775         cost the same as AVX128 ones.
4777 2018-10-08  Paul Koning  <ni1d@arrl.net>
4779         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
4780         (expand_block_move): New function.
4781         * config/pdp11/pdp11.c (output_block_move): Remove.
4782         (expand_block_move): New function.
4783         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
4784         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
4785         (*movmemhi1): Remove.
4787 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
4789         * config/s390/2827.md: Increase latencies for some FP instructions.
4791 2018-10-08  Richard Biener  <rguenther@suse.de>
4793         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4794         Open a dump scope.
4795         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
4796         * tree-vectorizer.h (dump_stmt_cost): Adjust.
4797         (add_stmt_cost): Dump return value of the hook.
4799 2018-10-08  Richard Biener  <rguenther@suse.de>
4801         PR tree-optimization/63155
4802         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
4803         (ssa_propagation_engine::ssa_propagate): Remove redundant
4804         bitmap bit clearing.
4806 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
4808         PR rtl-optimization/86939
4809         PR rtl-optimization/87479
4810         * ira.h (non_conflicting_reg_copy_p): New prototype.
4811         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
4812         (make_hard_regno_dead): Don't add conflicts for register
4813         ignore_reg_for_conflicts.
4814         (make_object_dead): Likewise.
4815         (non_conflicting_reg_copy_p): New function.
4816         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
4817         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
4818         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
4819         (make_hard_regno_dead): Don't add conflicts for register
4820         ignore_reg_for_conflicts.  Remove special conflict handling of
4821         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
4822         check_pic_pseudo_p and update callers.
4823         (mark_pseudo_dead): Don't add conflicts for register
4824         ignore_reg_for_conflicts.
4825         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
4827 2018-10-05  Andrew Waterman  <andrew@sifive.com>
4828             Jim Wilson  <jimw@sifive.com>
4830         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
4831         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
4832         new pattern using HONOR_SNANS that emits one extra instruction.
4834 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
4836         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
4837         patterns): Merge SI and DI patterns to a GPR pattern.
4838         (unnamed define_insn and define_split for record form of that): Merge
4839         to a single define_insn_and_split pattern.
4841 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
4843         PR c++/56856
4844         * input.c (expand_location_to_spelling_point): Add param "aspect"
4845         and use rather than hardcoding LOCATION_ASPECT_CARET.
4846         (get_substring_ranges_for_loc): Handle the case of a single token
4847         within a macro expansion.
4848         * input.h (expand_location_to_spelling_point): Add "aspect" param,
4849         defaulting to LOCATION_ASPECT_CARET.
4851 2018-10-05  Paul Koning  <ni1d@arrl.net>
4853         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
4854         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
4855         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
4856         (pdp11_guard_type): New function.
4858 2018-10-05  Paul Koning  <ni1d@arrl.net>
4860         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
4861         * config/pdp11/pdp11.opt (mfloat32): Remove.
4862         (mfloat64): Remove.
4863         * doc/invoke.texi (pdp11 -mfloat32): Remove:
4864         (pdp11 -mfloat64): Remove.
4866 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4868         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
4869         (*cmp<mode>_cc_i387): Ditto.
4870         (*cmpu<mode>_cc_i387): Ditto.
4871         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
4872         * config/i386/i386.c (ix86_expand_fp_compare): Remove
4873         "scratch" argument.
4874         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
4875         Emit x86_sahf_1 pattern.
4876         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
4877         (ix86_expand_carry_flag_compare): Ditto.
4879 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4881         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
4882         to reg_or_0_operand.  Add "C" constraint.
4883         (*cmpxf_cc_i387): Ditto.
4884         (*cmp<mode>_i387): Change operand 2 predicate
4885         to nonimm_or_0_operand.  Add "C" constraint.
4886         (*cmp<mode>_cc_i387): Ditto.
4887         (*cmp<mode>_0_i387): Remove insn pattern.
4888         (*cmp<mode>_0_cc_i387): Ditto.
4890 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4892         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
4893         * config/i386/predicates.md (nonimm_or_0_operand): Rename
4894         from vector_move_operand.  Update all uses.
4896 2018-10-05  Martin Sebor  <msebor@redhat.com>
4898         PR tree-optimization/87490
4899         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
4900         consistently.
4902 2018-10-05  Richard Biener  <rguenther@suse.de>
4904         PR tree-optimization/63155
4905         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
4906         vertical space in dumpfiles.
4907         * tree-ssa-propagate.h
4908         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
4909         * tree-ssa-propagate.c (cfg_blocks_back): New global.
4910         (ssa_edge_worklist_back): Likewise.
4911         (curr_order): Likewise.
4912         (cfg_blocks_get): Remove abstraction.
4913         (cfg_blocks_add): Likewise.
4914         (cfg_blocks_empty_p): Likewise.
4915         (add_ssa_edge): Add to current or next worklist based on
4916         RPO index.
4917         (add_control_edge): Likewise.
4918         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
4919         into ...
4920         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
4921         iteration from CFG and SSA edge worklist so we process
4922         everything in RPO order, prioritizing forward progress
4923         over iteration.
4924         (ssa_prop_init): Allocate new worklists, do not dump
4925         immediate uses.
4926         (ssa_prop_fini): Free new worklists.
4928 2018-10-05  Richard Biener  <rguenther@suse.de>
4930         * tree-core.h (tree_block::abstract_flag): Remove.
4931         (tree_block::block_num): Make full 32bits.
4932         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
4933         * tree.h (BLOCK_ABSTRACT): Remove.
4934         * dwarf2out.c (gen_lexical_block_die): Remove dead code
4935         resulting from BLOCK_ABSTRACT being always false.
4936         (gen_inlined_subroutine_die): Likewise.
4937         (gen_block_die): Likewise.
4938         * tree.c (block_ultimate_origin): Likewise.
4939         * tree-pretty-print.c (dump_block_node): Remove code dealing
4940         with BLOCK_ABSTRACT.
4941         * tree-ssa-live.c (dump_scope_block): Likewise.
4942         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
4943         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
4945 2018-10-05   Richard Biener  <rguenther@suse.de>
4947         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
4948         is asked for initialize mode to the component mode of the
4949         vector type.
4951 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
4953         PR target/87522
4954         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
4955         assembler for -mavx.
4956         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
4958 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
4960         PR target/87509
4961         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
4962         RS6000_BTM_DFP.
4963         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
4964         to be DImode.  When using mffscrn, force the operand to a register.
4966 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
4968         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
4969         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
4970         X87MODEF mode iterator.
4971         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
4972         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
4973         X87MODEF mode iterator.
4975 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
4977         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
4978         -Wno-prio-ctor-dtor.
4980 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
4982         * Makefile.in (OBJS): Add opt-problem.o.
4983         * dump-context.h: Include "selftest.h.
4984         (selftest::temp_dump_context): New forward decl.
4985         (class dump_context): Make friend of class
4986         selftest::temp_dump_context.
4987         (dump_context::dump_loc_immediate): New decl.
4988         (class dump_pretty_printer): Move here from dumpfile.c.
4989         (class temp_dump_context): Move to namespace selftest.
4990         (temp_dump_context::temp_dump_context): Add param
4991         "forcibly_enable_dumping".
4992         (selftest::verify_dumped_text):
4993         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
4994         (selftest::verify_item):
4995         (ASSERT_IS_TEXT): Move here from dumpfile.c.
4996         (ASSERT_IS_TREE): Likewise.
4997         (ASSERT_IS_GIMPLE): Likewise.
4998         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
4999         to...
5000         (dump_context::dump_loc_immediate): ...this new function.
5001         (class dump_pretty_printer): Move to dump-context.h.
5002         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
5003         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
5004         (temp_dump_context::temp_dump_context): Move to "selftest"
5005         namespace.  Add param "forcibly_enable_dumping", and use it to
5006         conditionalize the use of m_pp;
5007         (selftest::verify_dumped_text): Make non-static.
5008         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
5009         (selftest::verify_item): Make non-static.
5010         (ASSERT_IS_TEXT): Move to dump-context.h.
5011         (ASSERT_IS_TREE): Likewise.
5012         (ASSERT_IS_GIMPLE): Likewise.
5013         (selftest::test_capture_of_dump_calls): Pass "true" for new
5014         param of temp_dump_context.
5015         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
5016         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
5017         TDF_COMPARE_DEBUG.
5018         * opt-problem.cc: New file.
5019         * opt-problem.h: New file.
5020         * optinfo-emit-json.cc
5021         (selftest::test_building_json_from_dump_calls): Pass "true" for
5022         new param of temp_dump_context.
5023         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
5024         (optinfo::emit_for_opt_problem): New function.
5025         (optinfo::emit): Clarity which emit_item is used.
5026         * optinfo.h (optinfo::get_dump_location): New accessor.
5027         (optinfo::emit_for_opt_problem): New decl.
5028         (optinfo::emit): Make const.
5029         * selftest-run-tests.c (selftest::run_tests): Call
5030         selftest::opt_problem_cc_tests.
5031         * selftest.h (selftest::opt_problem_cc_tests): New decl.
5032         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
5033         bool to opt_result, converting fprintf messages to
5034         opt_result::failure_at calls.  Add "stmt" param for use by the
5035         failure_at calls.
5036         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
5037         (runtime_alias_check_p): Convert return type from bool to
5038         opt_result, converting dump_printf calls to
5039         opt_result::failure_at, using the statement DDR_A for their
5040         location.
5041         (find_data_references_in_stmt): Convert return type from bool to
5042         opt_result, converting "return false" to opt_result::failure_at
5043         with a new message.
5044         * tree-data-ref.h: Include "opt-problem.h".
5045         (dr_analyze_innermost): Convert return type from bool to opt_result,
5046         and add a const gimple * param.
5047         (find_data_references_in_stmt): Convert return type from bool to
5048         opt_result.
5049         (runtime_alias_check_p): Likewise.
5050         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
5051         dr_analyze_innermost.
5052         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
5053         Convert return type from bool to opt_result, adding a message for
5054         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
5055         (vect_analyze_data_ref_dependence): Convert return type from bool
5056         to opt_result.  Change sense of return type from "false"
5057         effectively meaning "no problems" to "false" meaning a problem,
5058         so that "return false" becomes "return opt_result::success".
5059         Convert "return true" calls to opt_result::failure_at, using
5060         the location of statement A rather than vect_location.
5061         (vect_analyze_data_ref_dependences): Convert return type from bool
5062         to opt_result.
5063         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
5064         calls to opt_result::failure_at, using the stmt location rather
5065         than vect_location.
5066         (vect_verify_datarefs_alignment): Convert return type from bool
5067         to opt_result.
5068         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
5069         into multiple more-tightly-scoped copies.
5070         (vect_analyze_data_refs_alignment): Convert return type from bool
5071         to opt_result.
5072         (vect_analyze_data_ref_accesses): Likewise, converting a
5073         "return false" to a "return opt_result::failure_at", adding a
5074         new message.
5075         (vect_prune_runtime_alias_test_list): Convert return type from
5076         bool to opt_result, converting dump_printf_loc to
5077         opt_result::failure_at.  Add a %G to show the pertinent statement,
5078         and use the stmt's location rather than vect_location.
5079         (vect_find_stmt_data_reference): Convert return type from
5080         bool to opt_result, converting dump_printf_loc to
5081         opt_result::failure_at, using stmt's location.
5082         (vect_analyze_data_refs):  Convert return type from bool to
5083         opt_result.  Convert "return false" to "return
5084         opt_result::failure_at", adding messages as needed.
5085         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
5086         type from bool to opt_result.
5087         (vect_determine_vf_for_stmt): Likewise.
5088         (vect_determine_vectorization_factor): Likewise, converting
5089         dump_printf_loc to opt_result::failure_at, using location of phi
5090         rather than vect_location.
5091         (vect_analyze_loop_form_1): Convert return type from bool to
5092         opt_result, converting dump_printf_loc calls, retaining the use of
5093         vect_location.
5094         (vect_analyze_loop_form): Convert return type from loop_vec_info
5095         to opt_loop_vec_info.
5096         (vect_analyze_loop_operations): Convert return type from bool to
5097         opt_result, converting dump_printf_loc calls, using the location
5098         of phi/stmt rather than vect_location where available.  Convert
5099         various "return false" to "return opt_result::failure_at" with
5100         "unsupported phi" messages.
5101         (vect_get_datarefs_in_loop): Convert return type from bool to
5102         opt_result.  Add a message for the
5103         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
5104         (vect_analyze_loop_2): Convert return type from bool to
5105         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
5106         each "goto again;", adding new messages where needed.
5107         Add "unsupported grouped {store|load}" messages.
5108         (vect_analyze_loop): Convert return type from loop_vec_info to
5109         opt_loop_vec_info.
5110         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
5111         bool to opt_result.
5112         * tree-vect-stmts.c (process_use): Likewise, converting
5113         dump_printf_loc call and using stmt location, rather than
5114         vect_location.
5115         (vect_mark_stmts_to_be_vectorized): Likeise.
5116         (vect_analyze_stmt): Likewise, adding a %G.
5117         (vect_get_vector_types_for_stmt): Convert return type from bool to
5118         opt_result, converting dump_printf_loc calls and using stmt
5119         location, rather than vect_location.
5120         (vect_get_mask_type_for_stmt): Convert return type from tree to
5121         opt_tree, converting dump_printf_loc calls and using stmt location.
5122         * tree-vectorizer.c: Include "opt-problem.h.
5123         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
5124         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
5125         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
5126         enabled, use it to report at the top level "couldn't vectorize
5127         loop" followed by the problem.
5128         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
5129         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
5130         to opt_result.
5131         (vect_analyze_stmt): Likewise.
5132         (vect_get_vector_types_for_stmt): Likewise.
5133         (tree vect_get_mask_type_for_stmt): Likewise.
5134         (vect_analyze_data_ref_dependences): Likewise.
5135         (vect_enhance_data_refs_alignment): Likewise.
5136         (vect_analyze_data_refs_alignment): Likewise.
5137         (vect_verify_datarefs_alignment): Likewise.
5138         (vect_analyze_data_ref_accesses): Likewise.
5139         (vect_prune_runtime_alias_test_list): Likewise.
5140         (vect_find_stmt_data_reference): Likewise.
5141         (vect_analyze_data_refs): Likewise.
5142         (vect_analyze_loop): Convert return type from loop_vec_info to
5143         opt_loop_vec_info.
5144         (vect_analyze_loop_form): Likewise.
5145         (vect_analyze_slp): Convert return type from bool to opt_result.
5147 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
5149         * doc/invoke.texi (-fopt-info): Document new "internals"
5150         sub-option.
5151         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
5152         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
5153         MSG_ALL_KINDS.
5154         (optinfo_verbosity_options): Add "internals".
5155         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
5156         (dump_context::apply_dump_filter_p): New member function.
5157         (dump_context::dump_loc): Use apply_dump_filter_p rather than
5158         explicitly masking the dump_kind.
5159         (dump_context::begin_scope): Increment the scope depth first.  Use
5160         apply_dump_filter_p rather than explicitly masking the dump_kind.
5161         (dump_context::emit_item): Use apply_dump_filter_p rather than
5162         explicitly masking the dump_kind.
5163         (dump_dec): Likewise.
5164         (dump_hex): Likewise.
5165         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
5166         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
5167         (opt_info_switch_p): Update handling of default
5168         MSG_OPTIMIZED_LOCATIONS to cope with default of
5169         MSG_PRIORITY_USER_FACING.
5170         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
5171         masking the dump_kind.
5172         (selftest::test_capture_of_dump_calls): Update test_dump_context
5173         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
5174         than MSG_ALL.  Generalize scope test to be run at all four
5175         combinations of with/without MSG_PRIORITY_USER_FACING and
5176         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
5177         for each of the two values.
5178         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
5179         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
5180         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
5181         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
5182         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
5183         with MSG_PRIORITY_*.
5184         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
5185         dump messages as MSG_PRIORITY_USER_FACING.
5186         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
5187         about the interaction with MSG_PRIORITY_*.
5189 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5191         * varasm.c (output_constant): Add new parameter merge_strings.
5192         Make strings properly zero terminated in merge string sections.
5193         (mergeable_string_section): Don't fail if the last char is non-zero.
5194         (assemble_variable_contents): Handle merge string sections.
5195         (assemble_variable): Likewise.
5196         (assemble_constant_contents): Likewise.
5197         (output_constant_def_contents): Likewise.
5198         (output_constructor_array_range,
5199         output_constructor_regular_field): Adjust call to output_constant.
5200         (output_object_block): Adjust call to assemble_constant_contents
5201         and assemble_variable_contents.
5203 2018-10-04  Martin Liska  <mliska@suse.cz>
5205         PR c/87483
5206         * cgraphunit.c (process_function_and_variable_attributes):
5207         Warn about a function with alias attribute and a body.
5209 2018-10-04  Martin Liska  <mliska@suse.cz>
5211         PR ipa/82625
5212         * multiple_target.c (redirect_to_specific_clone): New function.
5213         (ipa_target_clone): Use it.
5214         * tree-inline.c: Fix comment.
5216 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
5218         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
5219         fields.
5220         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
5221         (gcc::dump_manager::register_pass): New member function, adapted
5222         from loop body in gcc::pass_manager::register_pass, adding a
5223         call to update_dfi_for_opt_info.
5224         (gcc::dump_manager::opt_info_enable_passes): Store the
5225         -fopt-info options into the new fields.  Move the loop
5226         bodies into...
5227         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
5228         function.
5229         * dumpfile.h (struct opt_pass): New forward decl.
5230         (gcc::dump_manager::register_pass): New decl.
5231         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
5232         (class gcc::dump_manager): Add fields "m_optgroup_flags",
5233         "m_optinfo_flags", and "m_optinfo_filename".
5234         * passes.c (gcc::pass_manager::register_pass): Move all of the
5235         dump-handling code to gcc::dump_manager::register_pass.
5237 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
5239         PR rtl-optimization/87466
5240         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
5241         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
5242         * doc/tm.texi: Regenerate.
5243         * ira-lives.c (process_bb_node_lives): Use the new target hook.
5244         * lra-lives.c (process_bb_lives): Likewise.
5245         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
5246         Define.
5248 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
5250         * params.c (add_params): Fix initialization.
5252 2018-10-04  Martin Liska  <mliska@suse.cz>
5254         PR gcov-profile/84107
5255         * tree-profile.c (init_ic_make_global_vars):
5256         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
5257         Come up with new ic_tuple* variables.  Emit
5258         __gcov_indirect_call{,_topn} variables.
5259         (gimple_gen_ic_profiler): Access the variable
5260         and emit gimple.
5261         (gimple_gen_ic_func_profiler): Access
5262         __gcov_indirect_call.callee field.
5263         (gimple_init_gcov_profiler): Use ptr_type_node.
5264         * value-prof.c (gimple_ic): Use ptr_type_node.
5266 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5268         PR tree-optimization/85787
5269         * ipa-pure-const.c (malloc_candidate_p_1): Move most of
5270         malloc_candidate_p into this function and add support for
5271         detecting multiple phis.
5272         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
5274 2018-10-04  Martin Liska  <mliska@suse.cz>
5276         PR ipa/87491
5277         * ipa-inline.c (inline_to_all_callers_1):
5278         Call ultimate_alias_target for node being inlined.
5280 2018-10-03  Jeff Law  <law@redhat.com>
5282         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
5283         target's wchar_t.
5284         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
5285         * tree.h (get_typenode_from_name): Prototype.
5287 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
5289         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
5290         Change operand 2 predicate to nonimmediate_operand.
5291         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
5293 2018-10-03  Martin Sebor  <msebor@redhat.com>
5294             Jeff Law  <law@redhat.com>
5296         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
5297         initialize it.
5298         (get_string_length): Detect unterminated arrays.
5299         (format_string): Same.
5300         (format_directive): Warn about unterminated arrays.
5301         (handle_gimple_call): Mark statements with no_warning as needed.
5303 2018-10-03  Jim Wilson  <jimw@sifive.com>
5305         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
5306         also define __riscv_abi_rve.  Delete trailing white space.
5308 2018-10-03  Paul Koning  <ni1d@arrl.net>
5310         Enable LRA register allocator for PDP11.
5311         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
5312         (R): Likewise.
5313         (D): Likewise.
5314         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
5315         * config/pdp11/pdp11.opt (-mlra): New option.
5316         * doc/invoke.texi (PDP-11 Options): Document -mlra.
5318 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
5320         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
5321         (*<absneg:code>extend<mode>xf2): Ditto.
5323 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
5325         PR tree-optimization/87415
5326         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
5327         precision fields.
5329 2018-10-02  Jeff Law  <law@redhat.com>
5331         * gimple-fold.c (get_range_strlen): Only set *nonstr when
5332         an unterminated string is discovered.  Bubble up range
5333         even for unterminated strings.
5334         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
5335         indicates the string was not terminated via NONSTR.
5337 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
5339         * tree-vrp.c (extract_range_from_unary_expr): Special case all
5340         pointer conversions.
5341         Do not do anything special for anti-ranges.
5343 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
5345         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
5346         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
5348 2018-10-03  Martin Liska  <mliska@suse.cz>
5350         PR gcov-profile/86109
5351         * coverage.c (coverage_begin_function): Do not
5352         mark lambdas as artificial.
5353         * tree-core.h (struct GTY): Remove tm_clone_flag
5354         and introduce new lambda_function.
5355         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
5357 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
5359         PR target/87474
5360         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
5361         P8_VECTOR and VSX are enabled.
5363 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
5365         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
5366         0x3907 as CPU model number.
5368 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
5370         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
5371         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
5372         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
5373         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
5374         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
5375         * config/s390/s390.md: Likewise. Rename also the cpu attribute
5376         value from arch12 to z14.
5378 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
5380         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
5381         (isinfxf2): Ditto.
5382         (isinf<mode>2): Ditto.
5384 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
5386         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
5387         before emitting fxam.  Perform calculations in XFmode.
5389 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
5391         * match.pd (((X /[ex] A) +- B) * A): New transformation.
5393 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
5395         PR middle-end/87319
5396         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
5397         * tree.c (signed_or_unsigned_type_for): Handle complex.
5399 2018-10-02  Jeff Law  <law@redhat.com>
5401         * gimple-fold.c (get_range_strlen): Remove dead code.
5403 2018-10-02  Martin Sebor  <msebor@redhat.com>
5404             Jeff Law  <law@redhat.com>
5406         * builtins.c (unterminated_array): Add new arguments.
5407         If argument is not terminated, bubble up size and exact
5408         state to callers.
5409         (expand_builtin_strnlen): Detect, avoid expanding
5410         and diagnose unterminated arrays.
5411         (c_strlen): Fill in offset of start of unterminated strings.
5412         * builtins.h (unterminated_array): Update prototype.
5414 2018-10-02  Richard Biener  <rguenther@suse.de>
5416         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
5417         of haddv4df, first reduce to SSE width and exploit the fact
5418         that we only need element zero with the reduction result.
5419         (reduc_plus_scal_v2df): Likewise.
5421 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
5423         * dojump.h (do_jump): Delete.
5424         (do_jump_1): Likewise.
5425         (split_comparison): Move around.
5426         * dojump.c (do_jump): Make static.
5427         (do_jump_1): Likewise.
5428         (jumpifnot): Move around.
5429         (jumpifnot_1): Likewise.
5430         (jumpif): Likewise.
5431         (jumpif_1): Likewise.
5432         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
5434 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
5436         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
5437         insns in the delay slot and add_insn_after for the jump insn.
5439 2018-10-02  Richard Biener  <rguenther@suse.de>
5441         * tree-inline.c (expand_call_inline): Use the location of
5442         the callee declaration for the inline-entry marker.
5443         * final.c (notice_source_line): Remove special-casing of
5444         NOTE_INSN_INLINE_ENTRY.
5446 2018-10-01  Carl Love  <cel@us.ibm.com>
5448         PR 69431
5449         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
5450         (__builtin_mtfsb0): New.
5451         (__builtin_mtfsb1): New.
5452         ( __builtin_set_fpscr_rn): New.
5453         (__builtin_set_fpscr_drn): New.
5454         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
5455         (rs6000_expand_set_fpscr_rn_builtin): Add.
5456         (rs6000_expand_set_fpscr_drn_builtin): Add.
5457         (rs6000_expand_builtin): Add case statement entries for
5458         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
5459         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
5460         RS6000_BUILTIN_MFFSL.
5461         (rs6000_init_builtins): Add ftype initialization and def_builtin
5462         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
5463         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
5464         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
5465         rs6000_mffscdrn): Add define_insn.
5466         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
5467         * doc/extend.texi: Add documentation for the builtins.
5469 2018-10-01  Richard Biener  <rguenther@suse.de>
5471         PR tree-optimization/87465
5472         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
5473         causing branch miscounts.
5475 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5477         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
5478         aarch64_option_default_param):  New.
5479         (params.h): Include.
5480         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
5481         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
5482         stack-clash protection validation code.
5484 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5486         * params.c (validate_param): New.
5487         (add_params): Use it.
5488         (set_param_value): Refactor param validation into validate_param.
5489         (diagnostic.h): Include.
5490         * diagnostic.h (diagnostic_ready_p): New.
5492 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5494         * params.c (set_param_value):
5495         Add index of parameter being validated.
5496         * common/common-target.def (option_validate_param): New.
5497         * common/common-targhooks.h (default_option_validate_param): New.
5498         * common/common-targhooks.c (default_option_validate_param): New.
5499         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
5500         * doc/tm.texi: Regenerate.
5502 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5504         PR target/86486
5505         * config/aarch64/aarch64.c (aarch64_override_options_internal):
5506         Add validation for stack-clash parameters and set defaults.
5508 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5510         PR target/86486
5511         * configure.ac: Add stack-clash-protection-guard-size.
5512         * doc/install.texi: Document it.
5513         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
5514         * params.def: Update comment for guard-size.
5515         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
5516         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
5517         * configure: Regenerate.
5519 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5521         PR target/86486
5522         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
5523         STACK_DYNAMIC_OFFSET): New.
5524         * config/aarch64/aarch64.c (aarch64_layout_frame):
5525         Update outgoing args size.
5526         (aarch64_stack_clash_protection_alloca_probe_range,
5527         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
5529 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5531         PR target/86486
5532         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
5533         probe ranges.
5534         * target.def (stack_clash_protection_alloca_probe_range): New.
5535         (stack_clash_protection_final_dynamic_probe): Remove.
5536         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
5537         (default_stack_clash_protection_final_dynamic_probe): Remove.
5538         * targhooks.c: Likewise.
5539         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
5540         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
5541         * doc/tm.texi: Regenerate.
5543 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5545         PR target/86486
5546         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
5547         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
5548         aarch64_clamp_to_uimm12_shift): New.
5549         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
5550         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
5552 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5554         PR target/86486
5555         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
5557 2018-10-01  Jeff Law  <law@redhat.com>
5558             Richard Sandiford <richard.sandiford@linaro.org>
5559             Tamar Christina  <tamar.christina@arm.com>
5561         PR target/86486
5562         * config/aarch64/aarch64.md
5563         (probe_stack_range): Add k (SP) constraint.
5564         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
5565         STACK_CLASH_MAX_UNROLL_PAGES): New.
5566         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
5567         stack probes for stack clash.
5568         (aarch64_allocate_and_probe_stack_space): New.
5569         (aarch64_expand_prologue): Use it.
5570         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
5571         (aarch64_sub_sp): Add emit_move_imm optional param.
5573 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
5575         PR tree-optimization/87261
5576         * match.pd: Remove trailing whitespace.
5577         Add (x & y) | ~(x | y) -> ~(x ^ y),
5578         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
5580 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
5582         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
5583         constraints.
5584         (ashlsi3_insn): Update instruction constraints.
5585         (ashrsi3_insn): Likewise.
5586         (rotrsi3): Likewise.
5587         (add_shift): Likewise.
5588         * config/arc/constraints.md (Csz): New 32 bit constraint. It
5589         avoids placing in the limm field small constants which, otherwise,
5590         could end into a small instruction.
5592 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
5594         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
5595         destination register is not odd-even.
5596         (umaddsidi4_split): Likewise.
5598 2018-10-01  Richard Biener  <rguenther@suse.de>
5600         * tree-inline.c (expand_call_inline): Store origin of fn
5601         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
5602         * tree.c (block_ultimate_origin): Simplify and do some
5603         checking.
5605 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
5607         * config/i386/mmx.md (EMMS): New int iterator.
5608         (emms): New int attribute.
5609         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
5610         EMMS int iterator.  Explicitly declare clobbers.
5611         (mmx_emms): Remove expander.
5612         (mmx_femms): Ditto.
5613         * config/i386/predicates.md (emms_operation): Remove predicate.
5614         (vzeroall_pattern): New predicate.
5615         (vzeroupper_pattern): Rename from vzeroupper_operation.
5616         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
5617         vzeroupper_pattern and vzeroall_pattern predicates.
5619 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
5621         PR rtl-optimization/86939
5622         * ira-lives.c (make_hard_regno_born): Rename from this...
5623         (make_hard_regno_live): ... to this.  Remove update to conflict
5624         information.  Update function comment.
5625         (make_hard_regno_dead): Add conflict information update.  Update
5626         function comment.
5627         (make_object_born): Rename from this...
5628         (make_object_live): ... to this.  Remove update to conflict information.
5629         Update function comment.
5630         (make_object_dead):  Add conflict information update.  Update function
5631         comment.
5632         (mark_pseudo_regno_live): Call make_object_live.
5633         (mark_pseudo_regno_subword_live): Likewise.
5634         (mark_hard_reg_dead): Update function comment.
5635         (mark_hard_reg_live): Call make_hard_regno_live.
5636         (process_bb_node_lives): Likewise.
5637         * lra-lives.c (make_hard_regno_born): Rename from this...
5638         (make_hard_regno_live): ... to this.  Remove update to conflict
5639         information.  Remove now uneeded check_pic_pseudo_p argument.
5640         Update function comment.
5641         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
5642         to conflict information.  Update function comment.
5643         (mark_pseudo_live): Remove update to conflict information.  Update
5644         function comment.
5645         (mark_pseudo_dead): Add conflict information update.
5646         (mark_regno_live): Call make_hard_regno_live.
5647         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
5648         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
5650 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
5652         PR target/87370
5653         * config/i386/i386.c (construct_container): Use TImode for
5654         BLKmode values in 2 integer registers.
5656 2018-09-29  Jeff Law  <law@redhat.com>
5658         * builtins.c (unterminated_array): Pass in c_strlen_data * to
5659         c_strlen rather than just a tree *.
5660         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
5661         Update recursive calls appropriately.  If caller did not provide a
5662         suitable data pointer, create a local one.  When a non-terminated
5663         string is discovered, bubble up information about the string via the
5664         c_strlen_data object.
5665         * builtins.h (c_strlen): Update prototype.
5666         (c_strlen_data): New structure.
5667         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
5668         For a type 2 call, if c_strlen indicates a non-terminated string
5669         use the length of the non-terminated string.
5670         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
5672 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
5674         PR target/87467
5675         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
5676         __m512d type for __A argument rather than __m512.
5678 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
5680         * match.pd (simple_comparison): Don't optimize if either operand is
5681         a function pointer when target needs function pointer canonicalization.
5683 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5685         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
5686         power5 .. power9 to remove indirection.
5687         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
5688         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
5689         ASM_CPU_476_SPEC): Delete.
5690         (ASM_CPU_SPEC): Adjust.
5691         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
5692         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
5694 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5696         * config.in: Delete HAVE_AS_DCI.
5697         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
5698         * config/rs6000/rs6000.h: Ditto.
5699         * configure.ac: Delete HAVE_AS_DCI.
5700         * configure: Regenerate.
5702 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5704         * config.in (HAVE_AS_LWSYNC): Delete.
5705         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
5706         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
5707         do it as a .long .
5708         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
5709         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
5710         as a .long .
5711         * configure.ac: Delete HAVE_AS_LWSYNC.
5712         * configure: Regenerate.
5714 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
5715             Pierre-Marie de Rodat  <derodat@adacore.com>
5717         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
5718         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
5719         (cgraph_node::create_thunk): Add indirect_offset parameter.
5720         (thunk_adjust): Likewise.
5721         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
5722         and initialize the corresponding field with it.
5723         (cgraph_node::dump): Dump indirect_offset field.
5724         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
5725         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
5726         (thunk_adjust): Add indirect_offset parameter and deal with it.
5727         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
5728         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
5729         or if the thunk is external or local.  Fix formatting.  Do not chain
5730         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
5731         if any, in the GIMPLE representation.
5732         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
5733         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
5734         (input_node): Read indirect_offset field.
5735         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
5736         call to thunk_adjust.
5737         * tree-nested.c (struct nesting_info): Add thunk_p field.
5738         (create_nesting_tree): Set it.
5739         (convert_all_function_calls): Copy static chain from targets to thunks.
5740         (finalize_nesting_tree_1): Return early for thunks.
5741         (unnest_nesting_tree_1): Do not finalize thunks.
5742         (gimplify_all_functions): Do not gimplify thunks.
5744 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
5746         * opt-suggestions.c (option_proposer::build_option_suggestions):
5747         Release "option_values".
5749 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
5751         * coverage.c (get_coverage_counts): Convert problem-reporting dump
5752         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
5753         * dumpfile.c (kind_as_string): New function.
5754         (dump_loc): Rather than a hardcoded prefix of "note: ", use
5755         kind_as_string to vary the prefix based on dump_kind.
5756         (selftest::test_capture_of_dump_calls): Update for above.
5758 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
5760         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
5761         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
5763 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
5765         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
5766         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
5767         INVALID_REGNUM instead of FPSR_REG.
5768         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
5769         * config/i386/i386.md: Update comment of FP compares.
5770         (fldenv): Do not clobber FPSR_REG.
5772 2018-09-28  Richard Biener  <rguenther@suse.de>
5774         * tree.h (BLOCK_ORIGIN): New.
5775         * omp-expand.c (grid_expand_target_grid_body): Assign
5776         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
5777         * tree-inline.c (remap_block): Likewise.
5778         * auto-profile.c (get_function_decl_from_block): Simplify
5779         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
5780         * langhooks.c (lhd_print_error_function): Likewise.
5781         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
5782         Likewise.
5783         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5784         * tree.c (block_nonartificial_location): Likewise.
5785         (block_ultimate_origin): Likewise.
5786         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
5787         no longer needed LTO case.
5789 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
5790             Jan Hubicka  <jh@suse.cz>
5791             Martin Jambor  <mjambor@suse.cz>
5793         * simplify-rtx.c (simplify_merge_mask): New function.
5794         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
5795         same masks are used in op1 or op2.
5796         (test_vec_merge): New function.
5797         (test_vector_ops): Call test_vec_merge.
5799 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
5801         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
5802         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
5803         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
5804         (cypress_costs): Set it.
5805         (supersparc_costs): Likewise.
5806         (hypersparc_costs): Likewise.
5807         (leon_cost): Likewise.
5808         (leon3_costs): Likewise.
5809         (sparclet_costs): Likewise.
5810         (ultrasparc_costs): Likewise.
5811         (ultrasparc_costs): Likewise.
5812         (niagara_costs): Likewise.
5813         (niagara2_costs): Likewise.
5814         (niagara3_costs): Likewise.
5815         (niagara4_costs): Likewise.
5816         (niagara7_costs): Likewise.
5817         (m8_costs): Likewise.
5818         (TARGET_CAN_FOLLOW_JUMP): Define.
5819         (pass_work_around_errata::gate): Minor tweak.
5820         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
5821         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
5822         Automaitcally clear MASK_FSMULD mask for V7 processors.
5823         (sparc_can_follow_jump): New static function.
5824         (output_ubranch): Deal with CROSSING_JUMP_P.
5825         (sparc_use_sched_lookahead): Rewrite using switch statement.
5826         (sparc_issue_rate): Reorder.
5827         (sparc_branch_cost): New function.
5829 2018-09-27  Martin Sebor  <msebor@redhat.com>
5831         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
5832         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
5833         (int_fits_type_p): Same.
5835 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
5837         * config/i386/i386.md (FPCR_REG): Remove.
5838         (UNSPEC_FLDCW): Remove.
5839         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
5840         (x86_fldcw_1): Remove insn pattern.
5841         (fnstenv): Do not clobber FPCR_REG.
5842         (fldenv): Ditto.
5843         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
5844         (CALL_USED_REGISTERS): Ditto.
5845         (REG_ALLOC_ORDER): Ditto.
5846         (REG_CLASS_CONTENTS): Ditto.
5847         (HI_REGISTER_NAMES): Ditto.
5848         (ADDITIONAL_REGISTER_NAMES): Use defines instead
5849         of numerical constants.
5850         * config/i386/i386.c (regclass_map): Remove fpsr register.
5851         (dbx_register_map): Ditto.
5852         (dbx64_register_map): Ditto.
5853         (svr4_dbx_register_map): Ditto.
5854         (print_reg): Do not handle FPCR_REG.
5856 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
5858         PR target/87149
5859         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
5860         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
5861         Delete, always treat as true.
5862         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
5863         Ditto.  Simplify remaining code.
5864         * config/powerpcspe/powerpcspe.h: Ditto.
5865         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
5866         Simplify remaining code.
5867         (rs6000_expand_builtin): Ditto.
5868         * config/rs6000/rs6000.h: Ditto.
5869         * configure.ac: Ditto.
5870         * configure: Regenerate.
5872 2018-09-27  Martin Liska  <mliska@suse.cz>
5874         * coverage.c (get_coverage_counts): Revert the formatting
5875         of missing profile opt info.
5877 2018-09-27  Richard Biener  <rguenther@suse.de>
5879         PR debug/37801
5880         PR debug/87440
5881         * dwarf2out.c (set_block_origin_self): Do not mark outermost
5882         block as we do not output that.
5883         (gen_inlined_subroutine_die): Elide the originally outermost
5884         block, matching what we do for concrete instances.
5885         (decls_for_scope): Add parameter specifying whether to recurse
5886         to subblocks.
5888 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
5889             Tom de Vries  <tom@codesourcery.com>
5891         PR 82089
5893         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
5894         STORE_FLAG_VALUE == 1.
5896 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
5898         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
5899         constant definitions.
5900         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
5901         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
5902         ("speculation_barrier"): New expander definition.
5904 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
5906         PR gcov-profile/86957
5907         * common.opt: New warning option -Wmissing-profile.
5908         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
5909         * doc/invoke.texi: Document -Wmissing-profile.
5911 2018-09-26  Jim Wilson  <jimw@sifive.com>
5913         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
5914         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
5916 2018-09-26  Martin Sebor  <msebor@redhat.com>
5918         * tree.c (zerop): Change return type to bool.
5919         (integer_zerop, integer_onep, integer_each_onep): Same.
5920         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
5921         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
5922         (real_onep, real_minus_onep, chain_index): Same.
5923         (print_type_hash_statistics, type_list_equal): Same.
5924         * tree.h (zerop): Same.
5925         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
5926         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
5927         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
5928         (real_onep, real_minus_onep, chain_index): Same.
5929         (print_type_hash_statistics, type_list_equal): Same.
5931 2018-09-26  Jim Wilson  <jimw@sifive.com>
5933         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
5935 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
5937         PR target/87414
5938         * config/i386/i386.c: Include debug.h and dwarf2out.h.
5939         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
5940         call.
5942 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
5944         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
5946 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
5948         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
5949         and FP_SECOND_SSE_REGS.
5950         (REG_CLASS_NAMES): Ditto.
5951         (REG_CLASS_CONTENTS): Ditto.
5952         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
5953         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
5954         (ix86_preferred_output_reload_class): Ditto.
5955         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
5956         clobber constraint to "=&f".
5957         (fix_truncdi_i387): Ditto.
5958         (lrintxfdi2): Ditto.
5959         (fistdi2_<rounding>): Ditto.
5960         (fpremxf4_i387): Change "=u" constraint to "=f".
5961         (fprem1xf4_i387): Ditto.
5962         (sincosxf3): Ditto.
5963         (fptanxf4_i387): Ditto.
5964         (fxtractxf3_i387): Ditto.
5965         (fscalexf4_i387): Ditto.
5966         (atan2xf3): Change "u" constraint to "f".
5967         (fyl2xxf3_i387): Ditto.
5968         (fyl2xp1xf3_i387): Ditto.
5970 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
5972         PR target/87439
5973         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
5974         for removed I387_MASK_PM entity.
5977 2018-09-26  Jeff Law  <law@redhat.com>
5978         Revert
5979         2018-09-26  Alexey Neyman  <stilor@att.net>
5981         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
5982         headers are no longer pulled in by <isl/val.h>.
5984 2018-09-26  Richard Biener  <rguenther@suse.de>
5986         PR debug/87443
5987         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
5988         or concrete instance DIE to the tree.  Create abstract origin
5989         attributes also for concrete instances.
5991 2018-09-26  Alexey Neyman  <stilor@att.net>
5993         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
5994         headers are no longer pulled in by <isl/val.h>.
5996 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
5998         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
5999         Use new helper functions.
6000         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
6001         Use new helper functions.
6002         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
6003         aarch_mm_needs_release): New declarations.
6004         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
6005         aarch_mm_needs_release): New.
6007 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
6009         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
6010         (arm32_output_mi_thunk): Deal with long calls.
6012 2018-09-26  Richard Biener  <rguenther@suse.de>
6014         PR debug/87428
6015         PR debug/87362
6016         * tree-inline.c (expand_call_inline): When the location
6017         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
6018         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
6019         the inserted BLOCK to make inlined_function_outer_scope_p
6020         recognize it.
6021         * dwarf2out.c (add_call_src_coords_attributes): Do not add
6022         coords for reserved locations.
6024 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
6026         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
6027         (*call_indirect_nonlocal_sysv<mode>): Ditto.
6028         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
6029         (*sibcall_nonlocal_sysv<mode>): Ditto.
6030         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
6031         (<bd>_<mode>): Ditto.
6032         (<bd>tf_<mode>): Ditto.
6034 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
6036         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
6037         control string as a list of templates instead of as C code.
6038         (*altivec_movti): Ditto.
6039         * config/rs6000/darwin.md (movdf_low_di): Ditto.
6041 2018-09-25  Jim Wilson  <jimw@sifive.com>
6043         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
6044         when target symbol is weak.
6046 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6048         PR c/87387
6049         * builtins.c (unterminated_array): Simplify.
6050         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
6051         where pointer arithmetic is safe.
6053 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
6055         PR target/86987
6056         * config/rs6000/altivec.md (altivec_vspltb): Use
6057         const_0_to_15_operand instead of u5bit_cint_operand.
6058         (*altivec_vspltb_internal): Ditto.
6059         (altivec_vspltb_direct): Ditto.
6060         (altivec_vsplth): Use const_0_to_7_operand instead of
6061         u5bit_cint_operand.
6062         (*altivec_vsplth_internal): Ditto.
6063         (altivec_vsplth_direct): Ditto.
6064         (altivec_vspltw): Use const_0_to_3_operand instead of
6065         u5bit_cint_operand.
6066         (*altivec_vspltw_internal): Ditto.
6067         (altivec_vspltw_direct): Ditto.
6068         (altivec_vspltsf): Ditto.
6069         (*altivec_vspltsf_internal): Ditto.
6070         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
6071         various splats with the proper size immediate.  Reorder the various
6072         cases by ascending size of immediate, and put all such together.
6074 2018-09-25  Richard Biener  <rguenther@suse.de>
6076         PR debug/83941
6077         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
6078         GC-ification.
6079         (maybe_create_die_with_external_ref): Do not create
6080         DW_TAG_imported_unit here.
6081         (add_abstract_origin_attribute): Handle external BLOCK refs.
6082         (dwarf2out_abstract_function): Simplify LTO case.
6083         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
6084         rather than using maybe_create_die_with_external_ref.
6086 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
6088         PR target/71278
6089         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
6090         (frndintxf2_mask_pm_i387): Ditto.
6091         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
6092         Enable for !flag_trapping_math.
6093         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
6094         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
6095         Change operand 1 predicate to nonimmediate_operand.
6096         (attr "i387_cw"): Remove mask_pm.
6097         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
6098         (enum ix86_entity): Remove I387_MASK_PM.
6099         * config/i386/i386.c (ix86_i387_mode_needed): Do not
6100         handle I387_MASK_PM.
6101         (ix86_mode_needed): Ditto.
6102         (ix86_mode_after): Ditto.
6103         (ix86_mode_entry): Ditto.
6104         (ix86_mode_exit): Ditto.
6105         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
6107 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
6109         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
6110         to_update_switch_stmts to vNULL instead of calling create on them
6111         immediately.
6113 2018-09-25  Richard Biener  <rguenther@suse.de>
6115         PR tree-optimization/87402
6116         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
6117         (visit_phi): Re-instantiate handling of supposed to be VARYING
6118         but non-VARYING backedge value.
6120 2018-09-25  Richard Biener  <rguenther@suse.de>
6122         PR debug/83941
6123         * dwarf2out.c (struct sym_off_pair): New.
6124         (external_die_map): New global.
6125         (lookup_decl_die): When in LTO create DIEs lazily from the
6126         external_die_map.
6127         (lookup_block_die): New function, create DIEs lazily in LTO.
6128         (equate_block_to_die): New function.
6129         (dwarf2out_die_ref_for_decl): During WPA get the association
6130         from the external DIE map.
6131         (dwarf2out_register_external_die): Record mapping into the
6132         external DIE map.
6133         (maybe_create_die_with_external_ref): New function split out from
6134         DIE generation part of old dwarf2out_register_external_die.
6135         (add_abstract_origin_attribute): Do not return the DIE.  When
6136         in LTO reference externals directly.
6137         (dwarf2out_abstract_function): When in LTO ignore calls for
6138         decls with external DIEs (already present abstract instances).
6139         (gen_call_site_die): Adjust.
6140         (add_high_low_attributes): Likewise.
6141         (gen_lexical_block_die): Likewise.
6142         (gen_inlined_subroutine_die): Likewie.
6143         (gen_block_die): Likewise.
6144         (dwarf2out_inline_entry): Likewise.
6145         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
6146         DIEs.
6148 2018-09-25  Martin Liska  <mliska@suse.cz>
6150         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
6151         integers and not by a float value.
6153 2018-09-25  Martin Liska  <mliska@suse.cz>
6155         PR fortran/87394
6156         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
6157         instead of NULL.
6158         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
6159         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
6160         * godump.c (go_format_type): Remove extra parenthesis.
6162 2018-09-25  Martin Liska  <mliska@suse.cz>
6164         * alias.c (set_dest_equal_p): Remove unused function.
6165         * config/i386/i386.c (def_builtin_pure2): Likewise.
6166         * diagnostic-show-locus.c (class layout): Remove
6167         unused field.
6168         (layout::layout): Likewise here.
6169         * dump-context.h (class temp_dump_context): Likewise.
6170         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
6171         (add_AT_loclistsptr): Likewise.
6172         (add_AT_offset): Likewise.
6173         (get_AT_hi_pc): Likewise.
6174         (is_comdat_die): Likewise.
6175         (type_is_enum): Likewise.
6176         (ceiling): Likewise.
6177         (add_AT_vms_delta): Likewise.
6178         (is_class_die): Likewise.
6179         * edit-context.c (class line_event): Remove unused field.
6180         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
6181         unused function.
6182         * ipa-cp.c (ipa_get_vr_lat): Likewise.
6183         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
6184         (ok_for_base_p_nonstrict): Likewise.
6185         * tree-chrec.c (is_not_constant_evolution): Likewise.
6186         (chrec_fold_poly_cst): Likewise.
6187         * tree-if-conv.c (has_pred_critical_p): Likewise.
6188         * tree-ssa-coalesce.c (print_exprs): Likewise.
6189         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
6190         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
6191         * tree-vrp.c (value_ranges_intersect_p): Likewise.
6192         (value_range_nonnegative_p): Likewise.
6194 2018-09-25  Martin Liska  <mliska@suse.cz>
6196         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
6197         Do not handle "GNU Pascal".
6198         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
6199         Likewise.
6200         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
6201         from documentation. Likewise.
6202         * dbxout.c (dbxout_range_type): Likewise.
6203         * doc/cpp.texi: Likewise.
6204         * doc/extend.texi: Likewise.
6205         * doc/frontends.texi: Likewise.
6206         * doc/invoke.texi: Remove Pascal entry.
6207         * tree.def (CLEANUP_POINT_EXPR): Likewise.
6208         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
6210 2018-09-25  Martin Liska  <mliska@suse.cz>
6212         PR middle-end/86078
6213         * doc/invoke.texi: Document all parameters and remove default
6214         of the parameters.
6216 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
6218         PR bootstrap/87417
6219         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
6220         contains HOST_WIDE_INTs when computing its size.
6222 2018-09-24  Jim Wilson  <jimw@sifive.com>
6224         PR target/87391
6225         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
6226         not TARGET_RVE.
6227         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
6229 2018-09-24  Andrew Pinski  <apinski@marvell.com>
6231         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
6232         access prev before checking it for NULLness in the
6233         AARCH64_FUSE_CMP_BRANCH case.
6235 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
6237         PR target/82699
6238         * config/i386/i386.c (rest_of_insert_endbranch): Set
6239         endbr_queued_at_entrance to true and don't insert ENDBR if
6240         x86_function_profiler will be called.
6241         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
6242         is true.
6243         * config/i386/i386.h (machine_function): Add
6244         endbr_queued_at_entrance.
6246 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
6248         * genattrtab.c (mk_attr_alt): Use alternative_mask.
6249         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
6250         types.
6251         (check_attr_test): Use alternative_mask.
6252         (get_attr_value): Likewise.
6253         (compute_alternative_mask): Use alternative_mask and XWINT.
6254         (make_alternative_compare): Use alternative_mask.
6255         (attr_alt_subset_p): Use XWINT.
6256         (attr_alt_subset_of_compl_p): Likewise.
6257         (attr_alt_intersection): Use alternative_mask and XWINT.
6258         (attr_alt_union): Likewise.
6259         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
6260         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
6261         (simplify_test_exp): Use alternative_mask and XWINT.
6262         (write_test_expr): Use alternative_mask and XWINT, adjust bit
6263         number calculation to support 64 bits.  Generate code that
6264         checks 64-bit masks.
6265         (main): Use alternative_mask.
6266         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
6268 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
6270         PR target/80080
6271         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
6272         RETURN+USE when returning via %r14.
6274 2018-09-24  Martin Liska  <mliska@suse.cz>
6276         * gcov.c (output_lines): Print colorization legend
6277         for both flag_use_colors and flag_use_hotness_colors.
6278         Reword the help.
6280 2018-09-24  Martin Liska  <mliska@suse.cz>
6282         * coverage.c (get_coverage_counts): Use warning_at
6283         with current_function_decl location. Use %qD in warning
6284         message.
6286 2018-09-24  Martin Liska  <mliska@suse.cz>
6288         * memory-block.h (memory_block_pool::release): Annotate with
6289         valgrind that the memory is not accessible.
6291 2018-09-24  Martin Liska  <mliska@suse.cz>
6293         PR sanitizer/85774
6294         * asan.c: Make asan_handled_variables extern.
6295         * asan.h: Likewise.
6296         * cfgexpand.c (expand_stack_vars): Make sure
6297         a representative is unpoison if another
6298         variable in the partition is handled by
6299         use-after-scope sanitization.
6301 2018-09-24  Richard Biener  <rguenther@suse.de>
6303         PR tree-optimization/63155
6304         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
6305         the worklist when the edge of the respective argument isn't
6306         executable.
6308 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
6310         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
6311         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
6312         (MASK_CLASS_P): Update for rename.
6313         (MAYBE_MASK_CLASS_P): Ditto.
6314         (REG_CLASS_NAMES): Update.
6315         (REG_CLASS_CONTENT): Update.
6316         * config/i386/i386.c (regclass_map): Update for MASK_REG
6317         and ALL_MASK_REGS rename.
6318         * config/i386/constraints.md (Yk): Update for rename.
6319         (k): Ditto.
6321 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
6323         * config/i386/i386.h (enum reg_class): Remove
6324         EVEX_SSE_REGS and MOD4_SSE_REGS.
6325         (REG_CLASS_NAMES): Update.
6326         (REG_CLASS_CONTENT): Update.
6327         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
6328         registers as ALL_SSE_REGS.
6329         (ix86_additional_allocno_class_p): Remove.
6330         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
6331         (ix86_register_priority): Lower priority of EVEX SSE registers.
6332         Use IN_RANGE macro where appropriate.
6333         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
6334         AVX-5124VNNIW checks.
6335         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
6336         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
6337         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
6338         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
6339         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
6340         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
6341         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
6342         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
6343         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
6344         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
6345         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
6346         * config/i386/constraints.md (Yh): Remove.
6348 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
6350         * config/i386/i386.c (regclass_map): Declare integer REX registers
6351         as GENERAL_REGS.
6353 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
6355         * doc/service.texi (Service): Switch the fsf.org link to https.
6357 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
6359         PR target/86798
6360         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6361         Define to speculation_safe_value_not_needed.
6363 2018-09-21  Florian Weimer  <fweimer@redhat.com>
6365         PR middle-end/81035
6366         * doc/extend.texi (Common Function Attributes): Mention that
6367         noreturn suppresses tail call optimization.
6369 2018-09-21  Jeff Law  <law@redhat.com>
6371         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
6372         vr_values::cleanup_edges_and_switches.
6373         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
6374         vr_values class.
6375         (identify_jump_threads): Remove EDGE_IGNORE handling.
6376         (execute_vrp): Move handling of to_remove_edges and
6377         to_update_switch_stmts into vr_values class member functions.
6378         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
6379         (to_update_switch_stmts): Likewise.
6380         * vr-values.c: Include cfghooks.h.
6381         (vr_values::vr_values): Initialize to_remove_edges and
6382         to_update_switch_stmts.
6383         (vr_values::~vr_values): Verify to_remove_edges and
6384         to_update_switch_stmts are empty.
6385         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
6386         (vr_values::cleanup_edges_and_switches): New member function.
6387         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
6388         function.  Add new data members.
6390 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
6392         PR tree-optimization/87309
6393         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
6394         calls with pflags and alt_flags.
6395         (selftest::test_capture_of_dump_calls): Add test of interaction of
6396         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
6398 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6400         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
6402 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6404         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
6406 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6408         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
6409         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
6410         (VXWORKS_LIBS_RTP): Minor reordering.
6412 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6414         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
6415         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
6417 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6419         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
6420         (PTRDIFF_TYPE): Likewise.
6422 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6424         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
6425         triplet, similar to support for VxWorks7.
6426         * config/vxworks-dummy.h: Provide a default definition
6427         of TARGET_VXWORKS64 to 0.
6429 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6431         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
6432         * config/vxworks-dummy.h: here.
6434 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6436         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
6438 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
6439             Bo Zhou  <zbo.zhou@hisilicon.com>
6441         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
6442         * config/aarch64/aarch64-tune.md: Regenerated.
6443         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
6444         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
6445         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
6447 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
6448             Julian Brown  <julian@codesourcery.com>
6450         * builtins.c (get_builtin_sync_mem): Handle address spaces.
6452 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
6454         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
6455         if the call takes a static chain.
6457 2018-09-21  Martin Liska  <mliska@suse.cz>
6459         * auto-profile.c (autofdo_source_profile::read): Do not
6460         set sum_all.
6461         (read_profile): Do not add working sets.
6462         (read_autofdo_file): Remove sum_all.
6463         (afdo_callsite_hot_enough_for_early_inline): Remove const
6464         qualifier.
6465         * coverage.c (struct counts_entry): Remove gcov_summary.
6466         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
6467         do not support GCOV_TAG_PROGRAM_SUMMARY.
6468         (get_coverage_counts): Remove summary and expected
6469         arguments.
6470         * coverage.h (get_coverage_counts): Likewise.
6471         * doc/gcov-dump.texi: Remove -w option.
6472         * gcov-dump.c (dump_working_sets): Remove.
6473         (main): Do not support '-w' option.
6474         (print_usage): Likewise.
6475         (tag_summary): Likewise.
6476         * gcov-io.c (gcov_write_summary): Do not dump
6477         histogram.
6478         (gcov_read_summary): Likewise.
6479         (gcov_histo_index): Remove.
6480         (gcov_histogram_merge): Likewise.
6481         (compute_working_sets): Likewise.
6482         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
6483         it not obsolete.
6484         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
6485         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
6486         (GCOV_HISTOGRAM_SIZE): Remove.
6487         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
6488         (struct gcov_summary): Simplify rapidly just
6489         to runs and sum_max fields.
6490         (gcov_histo_index): Remove.
6491         (NUM_GCOV_WORKING_SETS): Likewise.
6492         (compute_working_sets): Likewise.
6493         * gcov-tool.c (print_overlap_usage_message): Remove
6494         trailing empty line.
6495         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
6496         (output_lines): Remove program related line.
6497         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
6498         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
6499         histogram.
6500         (input_profile_summary): Do not read it.
6501         (merge_profile_summaries): And do not merge it.
6502         (input_symtab): Do not call removed function.
6503         * modulo-sched.c (sms_schedule): Do not print sum_max.
6504         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
6505         removed when histogram method was invented.
6506         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
6507         mode.
6508         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
6509         GCOV coding style.
6510         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
6511         and dump selected value.
6512         * profile.c (add_working_set): Remove.
6513         (get_working_sets): Likewise.
6514         (find_working_set): Likewise.
6515         (get_exec_counts): Do not work with working sets.
6516         (read_profile_edge_counts): Do not inform as sum_max is removed.
6517         (compute_branch_probabilities): Likewise.
6518         (compute_value_histograms): Remove argument for call of
6519         get_coverage_counts.
6520         * profile.h: Do not make gcov_summary const.
6522 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
6524         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
6526 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
6528         PR tree-optimization/86990
6529         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
6530         Check that the entire merged store group is made of constants only for
6531         overlapping stores.
6533 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
6535         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
6536         (VTABLE_VERIFICATION_SPEC): Likewise.
6537         (SANITIZER_EARLY_SPEC): Likewise.
6538         (SANITIZER_SPEC): Likewise.
6539         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
6540         * doc/invoke.texi (Link Options): Document -r.
6542 2018-09-20  Richard Biener <rguenther@suse.de>
6544         PR middle-end/87054
6545         * gimplify.c (gimplify_expr): Retain alignment of
6546         addressable lvalue in dereference.
6548 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
6550         PR bootstrap/87013
6551         * configure.ac: Check for .loc is_stmt support.
6552         * configure, config.in: Rebuilt.
6553         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
6554         if not supported.
6556 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
6558         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
6559         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
6560         -misel=no.
6562 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
6564         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
6565         VECTOR_OTHER.
6566         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
6567         case VECTOR_OTHER.
6569 2018-09-20  Marek Polacek  <polacek@redhat.com>
6571         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
6573 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
6575         PR tree-optimization/87288
6576         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
6577         into account when determining PEELING_FOR_NITERS.
6579 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
6581         PR tree-optimization/86877
6582         * tree-vect-loop.c (vect_analyze_loop_2): Call
6583         vect_verify_datarefs_alignment.
6585 2018-09-19  Marek Polacek  <polacek@redhat.com>
6587         * doc/invoke.texi: Document -Wclass-conversion.
6589 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
6591         * config/pa/pa.c (pa_adjust_priority): Delete.
6592         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
6594         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
6595         (atomic_storehi): Likewise.
6596         (atomic_storesi): Likewise.
6597         (atomic_loaddi): Restore compare and swap exchange loop code.
6599 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
6601         PR rtl-optimization/86902
6602         * combine.c (try_combine): When changing the CC mode used, don't change
6603         an unrelated mode in other_insn to that new CC mode.
6605 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
6607         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
6608         with %T in place of calls to dump_generic_expr.
6609         (prune_runtime_alias_test_list): Likewise.
6610         (create_runtime_alias_checks): Likewise.
6611         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
6612         (vect_analyze_data_ref_dependence): Likewise.
6613         (vect_slp_analyze_data_ref_dependence): Likewise.
6614         (vect_record_base_alignment): Likewise.  Use %G in place of call
6615         to dump_gimple_stmt.
6616         (vect_compute_data_ref_alignment): Likewise.
6617         (verify_data_ref_alignment): Likewise.
6618         (vect_find_same_alignment_drs): Likewise.
6619         (vect_analyze_group_access_1): Likewise.
6620         (vect_analyze_data_ref_accesses): Likewise.
6621         (dependence_distance_ge_vf): Likewise.
6622         (dump_lower_bound): Likewise.
6623         (vect_prune_runtime_alias_test_list): Likewise.
6624         (vect_find_stmt_data_reference): Likewise.
6625         (vect_analyze_data_refs): Likewise.
6626         (vect_create_addr_base_for_vector_ref): Likewise.
6627         (vect_create_data_ref_ptr): Likewise.
6628         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
6629         (vect_can_advance_ivs_p): Likewise.
6630         (vect_update_ivs_after_vectorizer): Likewise.
6631         (vect_gen_prolog_loop_niters): Likewise.
6632         (vect_prepare_for_masked_peels): Likewise.
6633         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
6634         (vect_determine_vectorization_factor): Likewise.
6635         (vect_is_simple_iv_evolution): Likewise.
6636         (vect_analyze_scalar_cycles_1): Likewise.
6637         (vect_analyze_loop_operations): Likewise.
6638         (report_vect_op): Likewise.
6639         (vect_is_slp_reduction): Likewise.
6640         (check_reduction_path): Likewise.
6641         (vect_is_simple_reduction): Likewise.
6642         (vect_create_epilog_for_reduction): Likewise.
6643         (vect_finalize_reduction:): Likewise.
6644         (vectorizable_induction): Likewise.
6645         (vect_transform_loop_stmt): Likewise.
6646         (vect_transform_loop): Likewise.
6647         (optimize_mask_stores): Likewise.
6648         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
6649         (vect_split_statement): Likewise.
6650         (vect_recog_over_widening_pattern): Likewise.
6651         (vect_recog_average_pattern): Likewise.
6652         (vect_determine_min_output_precision_1): Likewise.
6653         (vect_determine_precisions_from_range): Likewise.
6654         (vect_determine_precisions_from_users): Likewise.
6655         (vect_mark_pattern_stmts): Likewise.
6656         (vect_pattern_recog_1): Likewise.
6657         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
6658         (vect_record_max_nunits): Likewise.
6659         (vect_build_slp_tree_1): Likewise.
6660         (vect_build_slp_tree_2): Likewise.
6661         (vect_print_slp_tree): Likewise.
6662         (vect_analyze_slp_instance): Likewise.
6663         (vect_detect_hybrid_slp_stmts): Likewise.
6664         (vect_detect_hybrid_slp_1): Likewise.
6665         (vect_slp_analyze_operations): Likewise.
6666         (vect_slp_analyze_bb_1): Likewise.
6667         (vect_transform_slp_perm_load): Likewise.
6668         (vect_schedule_slp_instance): Likewise.
6669         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
6670         (vect_mark_stmts_to_be_vectorized): Likewise.
6671         (vect_init_vector_1): Likewise.
6672         (vect_get_vec_def_for_operand): Likewise.
6673         (vect_finish_stmt_generation_1): Likewise.
6674         (vect_check_load_store_mask): Likewise.
6675         (vectorizable_call): Likewise.
6676         (vectorizable_conversion): Likewise.
6677         (vectorizable_operation): Likewise.
6678         (vectorizable_load): Likewise.
6679         (vect_analyze_stmt): Likewise.
6680         (vect_is_simple_use): Likewise.
6681         (vect_get_vector_types_for_stmt): Likewise.
6682         (vect_get_mask_type_for_stmt): Likewise.
6683         * tree-vectorizer.c (increase_alignment): Likewise.
6685 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
6687         * doc/rtl.texi: Adjust vec_select description.
6688         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
6689         non-constant selectors.
6691 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
6693         * config/aarch64/aarch64-protos.h
6694         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
6695         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
6696         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
6697         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
6698         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
6699         (AARCH64_FL_PROFILE): Move index so flags are ordered.
6700         (AARCH64_ISA_RCPC8_4): New flag.
6701         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
6702         to aarch64_offset_9bit_signed_unscaled_p.
6703         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
6704         and use stlur.
6705         * config/aarch64/constraints.md (Ust): New constraint.
6706         * config/aarch64/predicates.md.
6707         (aarch64_9bit_offset_memory_operand): New predicate.
6708         (aarch64_rcpc_memory_operand): New predicate.
6710 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
6712         PR rtl-optimization/87361
6713         * rtlanal.c (nonzero_bits1): Revert accidental change.
6715 2018-09-19  Richard Biener  <rguenther@suse.de>
6717         PR tree-optimization/87349
6718         PR tree-optimization/87342
6719         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
6721 2018-09-18  Marek Polacek  <polacek@redhat.com>
6723         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
6724         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
6726 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
6728         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
6730 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
6732         PR rtl-optimization/86882
6733         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
6735 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
6737         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
6738         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
6740 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
6742         PR other/87353
6743         * doc/invoke.texi (Link Options): Fix formatting and grammar.
6745 2018-09-18  Richard Biener  <rguenther@suse.de>
6747         PR middle-end/63155
6748         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
6749         (compute_samebase_partition_bases): Likewise.
6750         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
6751         (gimple_can_coalesce_p): Simplify.
6753 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
6755         Handle a library implementation of ffs calling __builtin_ffs.
6756         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
6757         (mmix_init_libfuncs): New function: make __builtin_ffs expand
6758         to __ffsdi2.
6760 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
6762         * diagnostic-show-locus.c (class layout_range): Add field
6763         "m_original_idx".
6764         (layout_range::layout_range): Add "original_idx" param and use it
6765         to initialize new field.
6766         (make_range): Use 0 for original_idx.
6767         (layout::layout): Pass in index to calls to
6768         maybe_add_location_range.
6769         (layout::maybe_add_location_range): Add param "original_idx" and
6770         pass it on to layout_range.
6771         (layout::print_any_labels): Pass on range->m_original_idx to
6772         get_text call.
6773         (gcc_rich_location::add_location_if_nearby): Use 0 for
6774         original_idx.
6775         * gcc-rich-location.h (text_range_label::get_text): Update for new
6776         param.
6777         (range_label_for_type_mismatch::get_text): Likewise.
6779 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
6781         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
6783 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
6785         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
6786         format_string_diagnostic_t.
6787         (fmtwarn_n): Likewise.
6788         * substring-locations.c
6789         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
6790         (format_warning_n_va): Convert to...
6791         (format_string_diagnostic_t::emit_warning_n_va): ...this.
6792         (format_warning_va): Convert to...
6793         (format_string_diagnostic_t::emit_warning_va): ...this.
6794         (format_warning_at_substring): Convert to...
6795         (format_string_diagnostic_t::emit_warning): ...this.
6796         (format_warning_at_substring_n): Convert to...
6797         (format_string_diagnostic_t::emit_warning_n): ...this.
6798         * substring-locations.h (class format_string_diagnostic_t): New
6799         class.
6800         (format_warning_va): Convert to
6801         format_string_diagnostic_t::emit_warning_va.
6802         (format_warning_n_va): Convert to
6803         format_string_diagnostic_t::emit_warning_n_va.
6804         (format_warning_at_substring): Convert to
6805         format_string_diagnostic_t::emit_warning.
6806         (format_warning_at_substring_n): Convert to
6807         format_string_diagnostic_t::emit_warning_n.
6809 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
6810             Bernd Schmidt <bernds_cb1@t-online.de>
6812         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
6813         SImode args.
6815 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
6817         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
6818         operand 0 predicate to nonimmediate operand.
6819         (rint<mode>2_frndint): Remove insn pattern.
6820         (rint<mode>2): Change operand 1 predicate to general_operand.
6821         Extend operand 1 to XFmode and generate rintxf2 insn.
6822         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
6823         Do not use X87MODEF mode macro.
6824         (frndintxf2_<rounding>_i387): Rename from
6825         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
6826         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
6827         to XFmode and generate significandxf3 insn.
6829 2018-09-17  Richard Biener  <rguenther@suse.de>
6831         PR tree-optimization/87328
6832         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
6833         visiting unexecutable backedges when not iterating.
6834         (do_rpo_vn): Mark all edges not executable even when not
6835         iterating.
6837 2018-09-17  Martin Jambor  <mjambor@suse.cz>
6839         PR c/63886
6840         * doc/invoke.texi (Warning Options): Likewise.
6842 2018-09-17  Richard Biener  <rguenther@suse.de>
6844         PR tree-optimization/87301
6845         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
6846         clean EH info from leftover copy assignments.
6848 2018-09-17  Martin Liska  <mliska@suse.cz>
6850         PR gcov-profile/85871
6851         * gcov.c (output_intermediate_file): Fix out of bounds
6852         access.
6854 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
6856         * config/arc/arc.c: Object attributes for core4 not reflected
6857         correctly.
6858         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
6859         core3).
6861 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
6863         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
6865 2018-09-17  Martin Liska  <mliska@suse.cz>
6867         * doc/gcov.texi: Document new option --use-hotness-colors.
6868         * gcov.c (struct source_info): Declare new field.
6869         (source_info::source_info): Set default for maximum_count.
6870         (print_usage): Add new -q option.
6871         (process_args): Process it.
6872         (accumulate_line_info): Save src->maximum_count.
6873         (output_line_beginning): Make color line number if
6874         flag_use_hotness_colors is set.
6875         (output_line_details): Pass default argument value.
6876         (output_lines): Pass src->maximum_count.
6878 2018-09-17  Martin Liska  <mliska@suse.cz>
6880         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
6881         Use processor_names table.
6882         * config/i386/i386.c (ix86_default_align): Use
6883         processor_cost_table for alignment values.
6884         (ix86_option_override_internal): Use processor_names.
6885         (ix86_function_specific_print): Likewise.
6886         * config/i386/i386.h (struct processor_costs):
6887         Add alignment values.
6888         (struct ptt): Remove and replace with const char *.
6889         * config/i386/x86-tune-costs.h (struct processor_costs):
6890         Declare default alignments for all costs.
6892 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
6894         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
6895         symbolics or VR_VARYING ranges for ABS_EXPR.
6896         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
6897         when range will wrap.
6899 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
6901         PR middle-end/86864
6902         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
6903         before and after a JUMP_TABLE_DATA.
6905 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
6907         PR middle-end/87188
6908         * dojump.c (do_compare_and_jump): Canonicalize function pointers
6909         when one operand is a function pointer.  Use POINTER_TYPE_P and
6910         FUNC_OR_METHOD_TYPE_P.
6911         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
6912         * fold-const.c (build_range_check): Likewise.
6913         * match.pd (simple_comparison): Likewise.
6915 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
6917         PR c/82967
6918         * spellcheck.c (get_edit_distance_cutoff): New function.
6919         (selftest::test_edit_distance_unit_test_oneway): Rename to...
6920         (selftest::test_get_edit_distance_one_way): ...this.
6921         (selftest::test_get_edit_distance_unit): Rename to...
6922         (selftest::test_get_edit_distance_both_ways): ...this.
6923         (selftest::test_edit_distances): Move tests to this new function,
6924         and test some more pairs of strings.  Update for above renaming.
6925         (selftest::get_old_cutoff): New function.
6926         (selftest::test_get_edit_distance_cutoff): New function.
6927         (selftest::assert_suggested_for): New function.
6928         (ASSERT_SUGGESTED_FOR): New macro.
6929         (selftest::assert_not_suggested_for): New function.
6930         (ASSERT_NOT_SUGGESTED_FOR): New macro.
6931         (selftest::test_suggestions): New function.
6932         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
6933         tests to selftest::test_edit_distances and call it.  Add calls to
6934         selftest::test_get_edit_distance_cutoff and
6935         selftest::test_suggestions.
6936         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
6937         (best_match::consider): Replace hard-coded cutoff calculation with
6938         a call to...
6939         (best_match::get_cutoff): New declaration.
6940         (best_match::get_best_meaningful_candidate): Likewise.
6942 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6944         * builtins.c (fold_builtin_strlen): Remove TODO comment.
6946 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6948         revert:
6949         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6951         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
6952         terminated string literal.
6954 2018-09-14  Martin Sebor  <msebor@redhat.com>
6956         * builtins.c (unterminated_array): Handle ARRAY_REF.
6957         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
6958         * builtins.h (unterminated_array): Declare extern.
6959         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
6960         arrays.
6961         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
6962         calls.
6964 2018-09-14  Martin Sebor  <msebor@redhat.com>
6965             Jeff Law  <law@redhat.com>
6967         * builtins.c (unterminated_array): New.
6968         (expand_builtin_strcpy): Adjust.
6969         (expand_builtin_strcpy_args): Detect unterminated arrays.
6970         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
6971         unterminated arrays.
6972         * gimple-fold.h (get_maxval_strlen): Add argument.
6973         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
6975         * gimple-fold.c (get_range_strlen): Add argument.
6976         (get_maxval_strlen): Adjust.
6977         * gimple-fold.h (get_range_strlen): Add argument.
6979 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
6981         * config/i386/movdirintrin.h: Fix copyright year.
6983 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
6985         * reg-stack.c: Include regs.h.
6986         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
6987         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
6988         FIRST_STACK_REG, not DFmode.
6989         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
6990         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
6991         (change stack): Default register mode to the reg_raw_mode of
6992         FIRST_STACK_REG, not DFmode.
6993         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
6994         (*swapxf): Rename from swapxf.
6996 2018-09-14  Carl Love  <cel@us.ibm.com>
6998         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
6999         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
7001 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
7003         PR target/87224
7004         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
7005         alternatives.
7007 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
7009         PR target/85628
7010         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
7012 2018-09-14  Jason Merrill  <jason@redhat.com>
7014         Fix --enable-gather-detailed-mem-stats.
7015         * hash-table.c (hash_table_usage): Change from variable to function.
7016         * hash-table.h: Adjust.
7017         * Makefile.in: Add missing dependencies on hash-table.h.
7019 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7021         PR tree-optimization/87259
7022         PR lto/87283
7023         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
7024         execute_cse_reciprocals_1 has tried transforming.
7026 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
7028         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
7029         VR_VARYING for PLUS/MINUS_EXPR.
7031 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
7033         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
7034         formatting.
7036 2018-09-14  Richard Biener  <rguenther@suse.de>
7038         PR middle-end/63155
7039         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
7040         bits for the merged partition.
7042 2018-09-13  Martin Sebor  <msebor@redhat.com>
7043             Bernd Edlinger  <bernd.edlinger@hotmail.de>
7045         * builtins.h (c_srlen): Add argument.
7046         * builtins.c (warn_string_no_nul): New function.
7047         (c_strlen): Add argument and use it.  Update recursive calls.
7048         Pass DECL argument to string_constant to get info on non
7049         terminated strings.  Update *NONSTR as needed.
7050         (fold_builtin_strlen): Add argument to calls to c_strlen.
7051         Warn for unterminated arrays.
7052         (warn_string_no_null): Add prototype.
7053         * expr.c (string_constant): Update arguments.  Update recursive
7054         calls appropriately.  Detect missing NUL terminator and outermost
7055         declaration its missing in.
7056         Improve checks for arrays with nonzero lower bound or elements
7057         that are not a single byte.  Simplify offset computation.
7058         Simplify checks for non-NUL terminated strings.
7059         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
7060         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
7062 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7064         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
7065         correctly.
7066         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
7067         argument.  Fix range checks.
7068         * fold-const.h (c_getstr): Adjust protoype.
7069         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
7070         string is constant but contains no NUL byte.
7072         * expr.c (string_constant): Adjust function comment.
7073         Remove bogus check for zero termination.
7075         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
7077         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
7078         (get_constant_size): Don't make STRING_CSTs larger than they are.
7079         (check_string_literal): New check function for STRING_CSTs.
7080         (output_constant): Use it.
7082 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
7084         PR target/86812
7085         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
7087 2018-09-13  Richard Biener  <rguenther@suse.de>
7089         PR tree-optimization/87263
7090         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
7091         (struct unwind_state): Add max_rpo field.
7092         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
7093         Compute max_rpo, the max RPO number a block can be backwards reached
7094         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
7095         separating it from the iterating mode.
7097 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
7099         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
7100         (rfs_decision): New scheduling decision.
7102 2018-09-13  Richard Biener  <rguenther@suse.de>
7104         PR bootstrap/87134
7105         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
7106         (vn_nary_op_insert_pieces_predicated): Do not write useless
7107         valid_dominated_by_p entry outside of the allocated storage.
7109 2018-09-13  Omar Sandoval  <osandov@osandov.com>
7110             Tom de Vries  <tdevries@suse.de>
7112         PR debug/86985
7113         * dwarf2out.c (is_c): New function.
7114         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
7116 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
7118         PR target/85628
7119         * config/aarch64/aarch64.md (*aarch64_bfxil):
7120         Define.
7121         * config/aarch64/constraints.md (Ulc): Define.
7122         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
7123         Define.
7124         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
7125         New function.
7127 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
7129         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
7130         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
7131         aarch64_layout_frame call.
7132         (aarch64_expand_epilogue): Likewise.
7133         (aarch64_initial_elimination_offset): Likewise.
7134         (aarch64_get_separate_components): Likewise.
7135         (aarch64_use_return_insn_p): Likewise.
7136         (aarch64_layout_frame): Remove unneeded check.
7138 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
7140         * configure.ac: Only append
7141         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
7142         gcc_config_arguments if it was never reconfigured or last reconfigure
7143         was with different arguments.
7144         * configure: Regenerated.
7146 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
7147             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7149         PR middle-end/87290
7150         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
7151         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
7153 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
7155         PR tree-optimization/87287
7156         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
7157         X % C == 0 to X % (unsigned) C == 0 optimization to ...
7158         * match.pd (X % C == 0): ... here.  New optimization.
7160 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
7162         PR middle-end/82853
7163         * expr.h (maybe_optimize_mod_cmp): Declare.
7164         * expr.c (mod_inv): New function.
7165         (maybe_optimize_mod_cmp): New function.
7166         (do_store_flag): Use it.
7167         * cfgexpand.c (expand_gimple_cond): Likewise.
7169 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
7170             Julian Brown  <julian@codesourcery.com>
7172         PR middle-end/86336
7173         * gimplify.c (gimplify_scan_omp_clauses): Set
7174         target_firstprivatize_array_bases in OpenACC parallel and kernels
7175         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
7176         OpenACC data regions.
7178 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
7180         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
7181         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
7183 2018-09-12  Richard Biener  <rguenther@suse.de>
7185         PR tree-optimization/87280
7186         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
7187         edge but unreachable target.
7188         (do_rpo_vn): For conservatively handling a PHI only mark
7189         the backedge executable but not the block reachable.
7191 2018-09-12  Richard Biener  <rguenther@suse.de>
7193         PR tree-optimization/87266
7194         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
7195         visited blocks.
7197 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
7199         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
7200         constants.
7201         ("trunc<BFP:mode><DFP_ALL:mode>2")
7202         ("trunc<DFP_ALL:mode><BFP:mode>2")
7203         ("extend<BFP:mode><DFP_ALL:mode>2")
7204         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
7205         according to the target operand type.
7207 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
7208             Andreas Krebbel  <krebbel@linux.ibm.com>
7210         PR tree-optimization/86844
7211         * gimple-ssa-store-merging.c
7212         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
7213         there are any overlapping stores in between them, make sure they are
7214         also coalesced or we give up completely.
7216 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
7218         PR middle-end/87248
7219         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
7220         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
7222 2018-09-12  Tom de Vries  <tdevries@suse.de>
7224         * common.opt (gdescribe-dies): Add option.
7225         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
7226         attribute for artifical and nameless decls.
7227         (dwarf2out_register_external_die): Add description attribute to
7228         external reference die.
7229         (add_desc_attribute): New functions.
7230         (gen_subprogram_die): Add description attribute to
7231         DW_TAG_call_site_parameter.
7232         * tree-pretty-print.c (print_generic_expr_to_str): New function.
7233         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
7234         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
7235         -gno-describe-dies.
7236         (@item -gdescribe-dies): Add.
7238 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
7240         * tree-vrp.c (vrp_shift_undefined_p): Remove.
7241         (extract_range_from_binary_expr_1: Call
7242         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
7243         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
7244         depend on sign.
7246 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
7248         * gimple-ssa-warn-alloca.c
7249         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
7250         field for ALLOCA_BOUND_*_LARGE.
7252 2018-09-11  Nathan Sidwell  <nathan@acm.org>
7254         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
7256 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
7258         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
7259         for clobbers.  Remove obsolete comment.
7261 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
7263         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
7264         mpxchk, mpxld and mpxst types.
7265         (define_attr length_immediate): Remove all processing of mpx types.
7266         (define_attr prefix_0f): Ditto.
7267         (define_attr memory): Ditto.
7269 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
7271         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
7272         (log<mode>2): Change operand 1 predicate to general_operand.
7273         Extend operand 1 to XFmode and generate logxf3 insn.
7274         (log10<mode>2): Change operand 1 predicate to general_operand.
7275         Extend operand 1 to XFmode and generate log10xf3 insn.
7276         (log2<mode>2): Change operand 1 predicate to general_operand.
7277         Extend operand 1 to XFmode and generate log2xf3 insn.
7278         (fyl2xp1_extend<mode>xf3_i387): Remove.
7279         (log1p<mode>2): Change operand 1 predicate to general_operand.
7280         Extend operand 1 to XFmode and generate log1pxf3 insn.
7281         (fxtract_extend<mode>xf3_i387): Remove.
7282         (logb<mode>2): Change operand 1 predicate to general_operand.
7283         Extend operand 1 to XFmode and generate logbxf3 insn.
7284         (ilogb<mode>2): Change operand 1 predicate to general_operand.
7285         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
7286         (significand<mode>2): Change operand 1 predicate to general_operand.
7287         Extend operand 1 to XFmode and generate significandxf3 insn.
7289 2018-09-11  Nathan Sidwell  <nathan@acm.org>
7291         * gcc.c (perror_with_name, pfatal_with_name): Delete.
7292         (load_specs): Use fatal_error.
7293         (DELETE_IF_ORDINARY, process_command): Use error.
7294         (execute, run_attempt): Use fatal_error.
7296 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
7298         * diagnostic-core.h (sorry_at): New prototype.
7299         * diagnostic.c (sorry_at): New function.
7301 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
7303         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
7304         by zero as VR_UNDEFINED.
7306 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
7308         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
7309         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
7310         (<sincos>mode2): New expander.
7311         (sincos_extend<mode>xf3_i387): Remove insn pattern.
7312         (sincos -> sin, cos splitters): Remove splitter patterns.
7313         (sincos<mode>3): Change operand 2 predicate to general_operand.
7314         Extend operand 2 to XFmode and generate sincosxf3 insn.
7315         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
7316         Change operand 3 predicate to const1_operand.
7317         (fptan_extend<mode>xf4_i387): Remove insn pattern.
7318         (tanxf2): Update operands in the call to fptanxf4_i387.
7319         (tan<mode>2): Change operand 1 predicate to general_operand.
7320         Extend operand 1 to XFmode and generate tanxf3 insn.
7321         (atan2xf3): Rename from *fpatanxf3_i387.
7322         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
7323         (atan2xf3): Remove expander.
7324         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
7325         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
7326         (atan<mode>2): Change operand 1 predicate to general_operand.
7327         Extend operand 1 to XFmode and generate atanxf3 insn.
7329 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
7331         * config/i386/i386.md (x87/SSE constant load splitter): Use
7332         memory_operand instead of nonimmediate_operand for input operand
7333         predicate.
7335 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
7337         * config/i386/i386.md (float partial SSE register stall splitter): Move
7338         splitter near its instruction pattern.
7339         (float_extend partial SSE register stall splitter): Ditto.
7340         (float_truncate partial SSE register stall splitter): Ditto.
7342 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
7344         PR target/86794
7345         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
7346         to speculation_safe_value_not_needed.
7348         PR target/85666
7349         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
7350         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
7351         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
7352         leaf_function_p, instead use has_hard_reg_initial_val.
7354 2018-09-09  Nathan Sidwell  <nathan@acm.org>
7356         * gcc.h (pfatal_with_name): Don't declare here.
7357         * gcc.c (pfatal_with_name): Make static.
7359 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
7361         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
7362         earlyclobber.
7364 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
7366         PR rtl-optimization/85458
7367         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
7368         priority hook to reduce the priority of EXPR.
7370 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
7372         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
7373         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
7374         with X87_ENABLE_FLOAT.  Remove preparation code.
7375         (*float<SWI48:mode><MODEF:mode>2): Rename from
7376         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
7377         math using "enabled" attribute.
7378         (*floatdi<MODEF:mode>2_i387): Rename from
7379         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
7380         enable for 32bit targets only.
7381         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
7382         splitter.
7383         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
7384         as operand 1 predicate.  Rewrite as define_insn_and_split.
7385         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
7387 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
7389         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
7390         to fallthru to FLOAT case.
7392 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
7394         PR target/86731
7395         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
7396         around folding of vec_sl to handle out of range shift values.
7398 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
7400         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
7401         Update callers to gen_fix_trunc<mode>_i387_fisttp
7402         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
7403         nonimmediate_operand.
7404         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
7405         and corresponding splitters.
7406         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
7407         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
7408         (fix_truncdi_i387_with_temp): Remove insn pattern
7409         and corresponding splitters.
7410         (fix_trunc<mode>_i387): Change operand 0 predicate to
7411         nonimmediate_operand.
7412         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
7413         and corresponding splitters.
7414         (*fistdi2_1): Remove.
7415         (fistdi2): Ditto.
7416         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
7417         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
7418         (*fist<mode>2_1): Remove.
7419         (fist<mode>2): Ditto.
7420         (fist<mode>2_with_temp): Remove insn pattern and corresponding
7421         splitters.
7422         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
7423         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
7424         (fistdi2_<rounding>): Change operand 0 predicate to
7425         nonimmediate_operand.
7426         (fistdi2_<rounding>_with_temp): Remove insn pattern
7427         and corresponding splitters.
7428         (fist<mode>2_<rounding>): Change operand 0 predicate to
7429         nonimmediate_operand.
7430         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
7431         and corresponding splitters.
7433         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
7435 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7437         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
7438         the init value.
7440 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
7442         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7443         early gimple folding of vec_splat().
7444         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
7445         * gimple-fold.h: Add an extern define for tree_vec_extract().
7447 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
7449         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
7450         wrappers around TREE_TYPE comparisons.
7452 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
7454         PR target/80080
7455         * config/s390/predicates.md: Add nonsym_memory_operand.
7456         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
7457         contains a SYMBOL_REF, load it into an intermediate pseudo.
7458         (s390_emit_compare_and_swap): Legitimize operand.
7459         * config/s390/s390.md: Use the new nonsym_memory_operand
7460         with UNSPECV_CAS patterns.
7462 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
7464         PR target/80080
7465         * config/s390/s390-passes.def: New file.
7466         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
7467         declaration.
7468         (make_pass_s390_early_mach): Add declaration.
7469         * config/s390/s390.c (make_pass_s390_early_mach):
7470         (s390_option_override): Remove dynamic registration.
7471         * config/s390/t-s390: Add s390-passes.def.
7473 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
7475         * config/s390/s390.c (s390_decompose_constant_pool_ref):
7476         Remove UNSPEC_LTREL_BASE check.
7477         (annotate_constant_pool_refs): Likewise.
7478         (find_constant_pool_ref): Likewise.
7479         (find_ltrel_base): Removed.
7480         (replace_ltrel_base): Removed.
7481         (s390_mainpool_finish): Remove replace_ltrel_base call.
7482         (s390_chunkify_start): Remove pending LTREL_BASE logic.
7483         (s390_chunkify_finish): Remove replace_ltrel_base call.
7484         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
7486 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
7488         PR target/86779
7489         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
7490         to speculation_safe_value_not_needed.
7492 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
7493             Bernd Schmidt  <bernds_cb1@t-online.de>
7495         * config/nvptx/nvptx-opts.h: New file.
7496         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
7497         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
7498         (ASM_SPEC): Define.
7499         (TARGET_SM35): New macro.
7500         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
7501         correct predicate.
7502         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
7503         values.
7504         (misa=): New option.
7505         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
7507 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
7509         * config/i386/i386.md (truncdfsf2): Remove expander.
7510         (truncdfsf2_with_temp): Ditto.
7511         (truncxf<mode>2): Ditto.
7512         (*truncdfsf_fast_mixed): Remove insn pattern.
7513         (*truncdfsf_fast_i387): Ditto.
7514         (*truncdfsf_mixed): Ditto.
7515         (*truncdfsf_i387): Ditto.
7516         (*truncdfsf2_i387_1): Ditto.
7517         (*truncxfsf2_mixed): Ditto.
7518         (*truncxfdf2_mixed): Ditto.
7519         (*truncxf<mode>2_i387_noop): Ditto. Update callers
7520         to call gen_truncxf<mode>2 instead.
7521         (*truncxf<mode>2_i387): Remove.
7522         (reg->reg splitters): Remove splitter pattern.
7523         (reg->mem splitters): Ditto.
7525         (truncdfsf2): New insn pattern.
7526         (truncxf<mode>2): Ditto.
7528 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7530         * tree-ssa-math-opts.c (is_mult_by): New function.
7531         (is_square_of): Use the above.
7532         (optimize_recip_sqrt): New function.
7533         (pass_cse_reciprocals::execute): Use the above.
7535 2018-09-05  Richard Biener  <rguenther@suse.de>
7537         PR bootstrap/87134
7538         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
7539         to zero-init the emplaced vec.
7541 2018-09-05  Martin Liska  <mliska@suse.cz>
7543         PR tree-optimization/87205
7544         * tree-switch-conversion.c (pass_lower_switch::execute):
7545         Group cases for switch statements.
7547 2018-09-05  Richard Biener  <rguenther@suse.de>
7549         PR tree-optimization/87217
7550         * tree-ssa-sccvn.c (vuse_valueize): New.
7551         (vn_reference_lookup_pieces): Use it.
7552         (vn_reference_lookup): Likewise.
7554 2018-09-05  Nathan Sidwell  <nathan@acm.org>
7556         PR c++/87137
7557         * stor-layout.c (place_field): Scan forwards to check last
7558         bitfield when ms_bitfield_placement is in effect.
7560 2018-09-05  Richard Biener  <rguenther@suse.de>
7562         PR bootstrap/87225
7563         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
7564         return.
7566 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
7567             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
7569         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
7570         * config.gcc (extra_objs): Build it.
7571         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
7572         Likewise.
7573         * config/aarch64/aarch64-passes.def
7574         (pass_tag_collision_avoidance): New pass.
7575         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
7576         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
7577         (aarch64_classify_address): Remove static qualifier.
7578         (aarch64_address_info, aarch64_address_type): Move to...
7579         * config/aarch64/aarch64-protos.h: ... here.
7580         (make_pass_tag_collision_avoidance): New function.
7581         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
7582         New tuning flag.
7584 2018-09-05  Martin Liska  <mliska@suse.cz>
7586         * doc/gcov.texi: Update documentation of humar
7587         readable mode.
7588         * gcov.c (format_count): Print one decimal place, it provides
7589         more fine number of situations like '1G' vs. '1.4G'.
7591 2018-09-05  Martin Liska  <mliska@suse.cz>
7593         PR target/87164
7594         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
7595         * optc-gen.awk: Allow 'Var' for Deprecated options in order
7596         to generate a MASK value.
7598 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
7600         PR debug/86593
7601         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
7602         if frame pointer isn't used.
7603         (compute_frame_pointer_to_fb_displacement): Likewise.
7605 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
7607         PR target/87198
7608         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
7609         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
7610         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
7611         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
7612         and OPTION_MASK_ISA_XSAVEC_UNSET.
7614 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
7616         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
7617         XOR operations in NAND case.
7619 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
7621         * wide-int-range.cc (wide_int_range_convert): New.
7622         * wide-int-range.h (wide_int_range_convert): New.
7623         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
7624         code into wide_int_range_convert.
7625         (extract_range_into_wide_ints): Do not munge anti range constants
7626         into the entire domain.  Just return the range back.
7628 2018-09-04  Martin Liska  <mliska@suse.cz>
7630         * genmatch.c (output_line_directive): Add new argument
7631         fnargs.
7632         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
7634 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
7636         * doc/invoke.texi (Option Summary): Add whitespace.
7638         * doc/invoke.texi (Option Summary): Add -Waligned-new.
7640 2018-09-04  Richard Biener  <rguenther@suse.de>
7642         PR tree-optimization/87211
7643         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
7644         backedge value we're supposed to treat as VARYING also number
7645         the PHI to VARYING in case it got a different value-number already.
7647 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
7649         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
7650         (extract_range_from_binary_expr_1): Do not call
7651         vrp_can_optimize_bit_op.
7652         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
7653         static.
7654         (wide_int_range_get_mask_and_bounds): New.
7655         (wide_int_range_optimize_bit_op): New.
7656         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
7657         (wide_int_range_bit_and): Same.
7658         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
7659         (wide_int_range_optimize_bit_op): New.
7660         (wide_int_range_get_mask_and_bounds): New.
7662 2018-09-04  Richard Biener  <rguenther@suse.de>
7664         PR tree-optimization/87176
7665         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
7666         variable.  When value-numbering a virtual PHI node make sure
7667         to not value-number to the backedge value.
7669 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
7671         * doc/extend.texi (Long Long, Hex Floats): Document support for
7672         long long and hex floats in more recent versions of ISO C++.
7674 2018-09-03  Richard Biener  <rguenther@suse.de>
7676         PR tree-optimization/87177
7677         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
7678         cleanup.
7680 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7682         * bb-reorder.c (edge_order): Convert to C-qsort-style
7683         tri-state comparator.
7684         (reorder_basic_blocks_simple): Change std::stable_sort to
7685         gcc_stablesort.
7687 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7689         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
7690         tri-state comparator.
7691         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
7693 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7695         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
7696         (mergesort): ... here as maximum count for using netsort.
7697         (gcc_qsort): Set nlim to 3 if stable sort is requested.
7698         (gcc_stablesort): New.
7699         * system.h (gcc_stablesort): Declare.
7701 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7703         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
7704         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
7705         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
7707 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
7709         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
7710         lxsdx and stxsdx alternatives.
7711         (*mov<mode>_hardfloat64): Ditto.
7712         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
7714 2018-09-03  Richard Biener  <rguenther@suse.de>
7716         PR tree-optimization/87200
7717         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
7718         simplify result.
7720 2018-09-03  Martin Liska  <mliska@suse.cz>
7722         PR tree-optimization/87201
7723         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
7724         Fix parenthesis in an expression.
7726 2018-09-03  Richard Biener  <rguenther@suse.de>
7728         PR tree-optimization/87197
7729         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
7730         visited.  CSE the VN_INFO hashtable lookup.
7732         PR tree-optimization/87169
7733         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
7734         iterating make sure there's no extra backedges from irreducible
7735         regions feeding the header.  Mark the destination block
7736         executable.
7738 2018-09-03  Martin Liska  <mliska@suse.cz>
7740         PR driver/83193
7741         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
7742         * common/common-targhooks.c (default_get_valid_option_values):
7743         New function.
7744         * common/common-targhooks.h (default_get_valid_option_values):
7745         Likewise.
7746         * common/config/i386/i386-common.c: Move processor_target_table
7747         from i386.c.
7748         (ix86_get_valid_option_values): New function.
7749         (TARGET_GET_VALID_OPTION_VALUES): New macro.
7750         * config/i386/i386.c (struct ptt): Move to i386-common.c.
7751         (PTA_*): Move all defined masks into i386-common.c.
7752         (ix86_function_specific_restore): Use new processor_cost_table.
7753         * config/i386/i386.h (struct ptt): Moved from i386.c.
7754         (struct pta): Likewise.
7755         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
7756         * doc/tm.texi.in: Likewise.
7757         * opt-suggestions.c (option_proposer::suggest_option):
7758         Pass prefix to build_option_suggestions.
7759         (option_proposer::get_completions): Likewise.
7760         (option_proposer::build_option_suggestions): Use the new target
7761         hook.
7762         * opts.c (struct option_help_tuple): New struct.
7763         (print_filtered_help): Use the new target hook.
7765 2018-09-03  Martin Liska  <mliska@suse.cz>
7767         PR middle-end/59521
7768         * predict.c (set_even_probabilities): Add likely_edges
7769         argument and handle cases where we have precisely one
7770         likely edge.
7771         (combine_predictions_for_bb): Catch also likely_edges.
7772         (tree_predict_by_opcode): Handle gswitch statements.
7773         * tree-cfg.h (find_case_label_for_value): New declaration.
7774         (find_taken_edge_switch_expr): Likewise.
7775         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
7776         Find pivot in decision tree based on probabily, not by number of
7777         nodes.
7779 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
7781         * doc/standards.texi (Standards): Update Objective-C reference.
7783 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7785         * doc/install.texi (Prerequisites): Update link for MPC.
7787 2018-09-01  Michael Matz  <matz@suse.de>
7789         PR tree-optimization/87074
7790         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
7791         PHIs for outer-loop uses.
7793 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7795         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
7796         * doc/invoke.texi (C Dialect Options): Ditto.
7798 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7800         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
7802 2018-08-31  Richard Biener  <rguenther@suse.de>
7804         PR tree-optimization/87168
7805         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
7806         (rpo_elim::eliminate_avail): When OP was not visited it must
7807         be available.
7809 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
7811         * tree-vrp.c (copy_value_range): Convert param "from" from
7812         "value_range *" to "const value_range *".
7813         (range_is_null): Likewise for param "vr".
7814         (range_int_cst_p): Likewise.
7815         (range_int_cst_singleton_p): Likewise.
7816         (symbolic_range_p): Likewise.
7817         (value_ranges_intersect_p): Likewise for both params.
7818         (value_range_nonnegative_p): Likewise for param "vr".
7819         (value_range_constant_singleton): Likewise.
7820         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
7821         (extract_range_into_wide_ints): Likewise for param "vr".
7822         (extract_range_from_multiplicative_op): Likewise for params "vr0"
7823         and "vr1".
7824         (vrp_can_optimize_bit_op): Likewise.
7825         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
7826         "vr1_".
7827         (extract_range_from_unary_expr): Likewise.
7828         (debug_value_range): Likewise for param "vr".
7829         (value_range::dump): Add "const" qualifier.
7830         (vrp_prop::check_array_ref): Convert local "vr" from
7831         "value_range *" to "const value_range *".
7832         (vrp_prop::check_mem_ref): Likewise.
7833         (vrp_prop::visit_stmt): Likewise for local "old_vr".
7834         (vrp_intersect_ranges_1): Likewise for param "vr_1".
7835         (vrp_intersect_ranges): Likewise.
7836         (simplify_stmt_for_jump_threading): Likewise for local "vr".
7837         (vrp_prop::vrp_finalize): Likewise.
7838         * tree-vrp.h (value_range::dump): Add "const" qualifier.
7839         (vrp_intersect_ranges): Add "const" qualifier to params as above.
7840         (extract_range_from_unary_expr): Likewise.
7841         (value_range_constant_singleton): Likewise.
7842         (symbolic_range_p): Likewise.
7843         (copy_value_range): Likewise.
7844         (extract_range_from_binary_expr_1): Likewise.
7845         (range_int_cst_p): Likewise.
7846         (vrp_set_zero_nonzero_bits): Likewise.
7847         (range_int_cst_singleton_p): Likewise.
7849 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
7851         * config/aarch64/arm_neon.h (vabsd_s64): New.
7852         (vnegd_s64): Likewise.
7854 2018-08-31  Martin Jambor  <mjambor@suse.cz>
7856         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
7858 2018-08-31  Martin Liska  <mliska@suse.cz>
7860         * ipa-icf.c (sem_item::add_type): Use
7861         sem_item::m_type_hash_cache.
7862         * ipa-icf.h: Move the cache from sem_item_optimizer
7863         to sem_item.
7865 2018-08-31  Nathan Sidwell  <nathan@acm.org>
7867         * doc/extend.texi (Backwards Compatibility): Remove implicit
7868         extern C leeway of () being (...).
7870 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7872         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
7874 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
7876         PR target/86684
7877         PR target/87149
7878         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
7880 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
7882         PR middle-end/87138
7883         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
7884         gen_int_mode.  Formatting fixes.
7886 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
7888         * target.def (custom_function_descriptors): Improve documentation.
7889         * doc/tm.texi.in (Trampolines): Expand discussion of function
7890         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
7891         beginning of the section.
7892         * doc/tm.texi: Regenerated.
7894 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
7896         * cfg.h (class auto_edge_flag): Spell out the template-id of the
7897         base class in the initializer list.  This is a workaround for
7898         building with older GCC.
7899         (class auto_bb_flag): Likewise.
7901 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
7903         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
7904         (altivec_vcmpequ<VI_char>_p): Remove star.
7905         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
7906         vector load modes.
7907         (expand_strncmp_vec_sequence): New function.
7908         (emit_final_str_compare_vec): New function.
7909         (expand_strn_compare): Add support for vector strncmp.
7910         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
7911         length specification to bytes.
7912         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
7913         (vcmpnezb_p): New pattern.
7914         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
7915         for option -mstring-compare-inline-limit.
7917 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
7919         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
7920         (PTA_SKYLAKE): Add PTA_AES.
7921         (PTA_GOLDMONT): Likewise.
7923 2018-08-29  Jan Hubicka  <jh@suse.cz>
7925         PR lto/86517
7926         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
7927         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
7929 2018-08-29  Jan Hubicka  <jh@suse.cz>
7931         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
7932         TYPE_STUB_DECL.
7933         (hash_tree): Do not visit TYPE_STUB_DECL.
7934         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
7935         stream TYPE_STUB_DECL.
7936         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
7937         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
7938         after free_lang_data.
7939         (type_in_anonymous_namespace_p): Likewise.
7941 2018-08-29  Jan Hubicka  <jh@suse.cz>
7943         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
7944         comment that it has to be even number.
7945         (class sreal): Change m_sig type to int32_t.
7946         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
7947         int64_t for temporary calculations.
7948         (sreal_verify_basics): Drop one bit from minimum and maximum.
7950 2018-08-30  Richard Biener  <rguenther@suse.de>
7952         PR tree-optimization/87147
7953         * tree-ssa-sccvn.c (SSA_VISITED): New function.
7954         (visit_phi): When the degenerate result is from the backedge and
7955         we didn't visit its definition yet drop to VARYING.
7956         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
7958 2018-08-29  Jan Hubicka  <jh@suse.cz>
7960         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
7961         DECL_VINDEX.
7962         (hash_tree): Likewise.
7964 2018-08-29  Jan Hubicka  <jh@suse.cz>
7966         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
7967         and TYPE_NEXT_REF_TO.
7969 2018-08-29  Jan Hubicka  <jh@suse.cz>
7971         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
7972         comment that it has to be even number.
7973         (class sreal): Change m_sig type to int32_t.
7974         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
7975         int64_t for temporary calculations.
7976         (sreal_verify_basics): Drop one bit from minimum and maximum.
7978 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
7980         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
7982 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
7984         PR middle-end/86995
7985         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
7986         if to_add is negative.
7988 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7990         PR middle-end/87053
7991         * builtins.c (c_strlen): Improve range checks.
7993 2018-08-29  Martin Sebor  <msebor@redhat.com>
7994             Jeff Law  <law@redhat.com>
7996         PR tree-optimization/86714
7997         PR tree-optimization/86711
7998         * builtins.c (c_strlen): Add arguments to call to string_constant.
7999         * expr.c (string_constant): Add argument.  Detect missing nul
8000         terminator and outermost declaration it's missing in.
8001         * expr.h (string_constant): Add argument.
8002         * fold-const.c (read_from_constant_string): Add arguments to call to
8003         string_constant.
8004         (c_getstr): Likewise.
8005         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
8006         to string_constant.
8007         * tree-ssa-strlen.c (get_stridx): Likewise.
8009 2018-08-29  Jan Hubicka  <jh@suse.cz>
8011         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
8012         Do not stream DECL_VINDEX.
8013         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
8014         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
8015         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
8017 2018-08-29  Richard Biener  <rguenther@suse.de>
8019         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
8020         virtual operands that are not default defs to honor region
8021         boundaries.
8022         (rpo_vn_valueize): Remove ineffective code here.
8024 2018-08-29  Richard Biener  <rguenther@suse.de>
8026         PR tree-optimization/87132
8027         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
8028         when skipping defs reachable over backedges.
8030 2018-08-29  Richard Biener  <rguenther@suse.de>
8032         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
8033         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
8034         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
8035         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
8036         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
8037         (vn_reference_lookup_3): Remove use of const_parms.
8038         (free_rpo_vn): Do not free const_parms.
8039         (do_rpo_vn): Do not call init_const_parms.
8040         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
8041         SSA_NAME_POINTS_TO_READONLY_MEMORY.
8042         (call_may_clobber_ref_p_1): Likewise.
8044 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
8046         PR other/86726
8047         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
8048         (-O): Ditto.
8049         (-ftree-scev-cprop): Document.
8051 2018-08-29  Jan Hubicka  <jh@suse.cz>
8053         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
8054         parameters.
8055         (sreal constructor): Update.
8056         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
8057         sreal:operator/): Update.
8059 2018-08-29  Martin Liska  <mliska@suse.cz>
8061         * tree-switch-conversion.c (switch_conversion::expand):
8062         Strenghten assumption about gswitch statements.
8064 2018-08-29  Richard Biener  <rguenther@suse.de>
8066         PR tree-optimization/87117
8067         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
8068         re-value-number released SSA VDEFs.
8070 2018-08-29  Richard Biener  <rguenther@suse.de>
8072         PR tree-optimization/87126
8073         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
8075 2018-08-28  Jim Wilson  <jimw@sifive.com>
8077         * config/riscv/pic.md: Rewrite.
8078         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
8079         invalid address.
8080         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
8081         (SOFTF, default_load, softload, softstore): New.
8083 2018-08-28  Jeff Law  <law@redhat.com>
8085         * fold-const.c (fold_binary_loc): Remove recently added assert.
8087 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
8089         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
8090         to OP parmeter of generated function.
8092 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
8094         PR tree-optimization/87009
8095         * match.pd: Add boolean optimizations.
8097 2018-08-28  Martin Sebor  <msebor@redhat.com>
8099         PR middle-end/86631
8100         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
8101         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
8102         (pass_walloca::gate): Use it.
8103         (alloca_call_type): Same.
8104         (pass_walloca::execute): Same.
8105         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
8107 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
8109         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
8110         GCC_VERSION for usage of "__gcc_dump_printf__" format from
8111         >= 3005 to >= 9000.
8113 2018-08-28  Richard Biener  <rguenther@suse.de>
8115         PR tree-optimization/87124
8116         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
8117         constants before looking up avail.
8119 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
8121         PR middle-end/87099
8122         * calls.c (maybe_warn_nonstring_arg): Punt early if
8123         warn_stringop_overflow is zero.  Don't call get_range_strlen
8124         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
8125         Swap comparison operands to have constants on rhs.  Only use
8126         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
8127         increment lenrng[0].
8129 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
8131         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
8132         use of tree_to_shwi.  Remove duplicated test for the size being
8133         a whole number of bytes.
8135 2018-08-28  Richard Biener  <rguenther@suse.de>
8137         PR tree-optimization/87117
8138         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
8139         Handle removed stmt without LHS (GIMPLE_NOP).
8141 2018-08-28  Richard Biener  <rguenther@suse.de>
8143         PR tree-optimization/87117
8144         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
8145         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
8147 2018-08-28  Richard Biener  <rguenther@suse.de>
8149         PR tree-optimization/87117
8150         * tree-ssa-pre.c (compute_avail): Do not make expressions
8151         with predicated values available.
8152         (get_expr_value_id): Assert we do not run into predicated value
8153         expressions.
8155 2018-08-28  Richard Biener  <rguenther@suse.de>
8157         PR tree-optimization/87117
8158         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
8159         get virtual operands.
8160         (get_expr_operands): Handle STRING_CST like other decls.
8162 2018-08-28  Martin Liska  <mliska@suse.cz>
8164         * tree.h: Update documentation of fndecl_built_in_p
8165         functions.
8168 2018-08-27  Jeff Law  <law@redhat.com>
8169         PR tree-optimization/87110
8170         * tree-ssa-dse.c (compute_trims): Handle non-constant
8171         TYPE_SIZE_UNIT.
8173 2018-08-27  Martin Sebor  <msebor@redhat.com>
8175         PR tree-optimization/86914
8176         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
8178 2018-08-27  Martin Sebor  <msebor@redhat.com>
8180         PR tree-optimization/87112
8181         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
8182         the type of the bound argument.
8184 2018-08-27  Jeff Law  <law@redhat.com>
8186         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
8187         type does not have a TYPE_SIZE_UNIT.
8189 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
8191         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
8192         with include of backend.h.
8194 2018-08-27  Richard Biener  <rguenther@suse.de>
8196         PR tree-optimization/86927
8197         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
8198         use const cond reduction code.
8200 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
8202         PR tree-optimization/85758
8203         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
8205 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
8207         PR c++/87091
8208         * diagnostic-show-locus.c (class layout_range): Update for
8209         conversion of show_caret_p to a tri-state.
8210         (layout_range::layout_range): Likewise.
8211         (make_range): Likewise.
8212         (layout::maybe_add_location_range): Likewise.
8213         (layout::should_print_annotation_line_p): Don't show annotation
8214         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
8215         (layout::get_state_at_point): Update for conversion of
8216         show_caret_p to a tri-state.  Bail out early for
8217         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
8218         underlining or source colorization.
8219         (gcc_rich_location::add_location_if_nearby): Update for conversion
8220         of show_caret_p to a tri-state.
8221         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
8222         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
8223         Likewise.
8224         (selftest::test_one_liner_labels): Likewise.
8225         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
8226         conversion of show_caret_p to a tri-state.
8227         * pretty-print.c (text_info::set_location): Likewise.
8228         * pretty-print.h (text_info::set_location): Likewise.
8229         * substring-locations.c (format_warning_n_va): Likewise.
8230         * tree-diagnostic.c (default_tree_printer): Likewise.
8231         * tree-pretty-print.c (newline_and_indent): Likewise.
8233 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
8235         PR c++/87091
8236         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
8237         line above for line-insertion fix-it hints.
8238         (selftest::test_fixit_insert_containing_newline): Update the
8239         expected results, and add a test with line-numbering enabled.
8241 2018-08-27  Martin Liska  <mliska@suse.cz>
8243         PR sanitizer/86962
8244         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
8245         params with DECL_HAS_VALUE_EXPR_P.
8247 2018-08-27  Martin Liska  <mliska@suse.cz>
8249         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
8250         selected expansion strategy.
8252 2018-08-27  Martin Liska  <mliska@suse.cz>
8254         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
8255         * builtins.c (is_builtin_fn): Likewise.
8256         * attribs.c (diag_attr_exclusions): Use new function
8257         fndecl_built_in_p and remove check for FUNCTION_DECL if
8258         possible.
8259         (builtin_mathfn_code): Likewise.
8260         (fold_builtin_expect): Likewise.
8261         (fold_call_expr): Likewise.
8262         (fold_builtin_call_array): Likewise.
8263         (fold_call_stmt): Likewise.
8264         (set_builtin_user_assembler_name): Likewise.
8265         (is_simple_builtin): Likewise.
8266         * calls.c (gimple_alloca_call_p): Likewise.
8267         (maybe_warn_nonstring_arg): Likewise.
8268         * cfgexpand.c (expand_call_stmt): Likewise.
8269         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
8270         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
8271         (cgraph_node::verify_node): Likewise.
8272         * cgraphclones.c (build_function_decl_skip_args): Likewise.
8273         (cgraph_node::create_clone): Likewise.
8274         * config/arm/arm.c (arm_insert_attributes): Likewise.
8275         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
8276         * dse.c (scan_insn): Likewise.
8277         * expr.c (expand_expr_real_1): Likewise.
8278         * fold-const.c (operand_equal_p): Likewise.
8279         (fold_binary_loc): Likewise.
8280         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
8281         * gimple-low.c (lower_stmt): Likewise.
8282         * gimple-pretty-print.c (dump_gimple_call): Likewise.
8283         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8284         Likewise.
8285         * gimple.c (gimple_build_call_from_tree): Likewise.
8286         (gimple_call_builtin_p): Likewise.
8287         (gimple_call_combined_fn): Likewise.
8288         * gimplify.c (gimplify_call_expr): Likewise.
8289         (gimple_boolify): Likewise.
8290         (gimplify_modify_expr): Likewise.
8291         (gimplify_addr_expr): Likewise.
8292         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
8293         * ipa-cp.c (determine_versionability): Likewise.
8294         * ipa-fnsummary.c (compute_fn_summary): Likewise.
8295         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
8296         * ipa-split.c (visit_bb): Likewise.
8297         (split_function): Likewise.
8298         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
8299         * lto-cgraph.c (input_node): Likewise.
8300         * lto-streamer-out.c (write_symbol): Likewise.
8301         * omp-low.c (setjmp_or_longjmp_p): Likewise.
8302         (lower_omp_1): Likewise.
8303         * predict.c (strip_predict_hints): Likewise.
8304         * print-tree.c (print_node): Likewise.
8305         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
8306         * trans-mem.c (is_tm_irrevocable): Likewise.
8307         (is_tm_load): Likewise.
8308         (is_tm_simple_load): Likewise.
8309         (is_tm_store): Likewise.
8310         (is_tm_simple_store): Likewise.
8311         (is_tm_abort): Likewise.
8312         (tm_region_init_1): Likewise.
8313         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
8314         * tree-cfg.c (verify_gimple_call): Likewise.
8315         (move_stmt_r): Likewise.
8316         (stmt_can_terminate_bb_p): Likewise.
8317         * tree-eh.c (lower_eh_constructs_2): Likewise.
8318         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
8319         * tree-inline.c (remap_gimple_stmt): Likewise.
8320         (copy_bb): Likewise.
8321         (estimate_num_insns): Likewise.
8322         (fold_marked_statements): Likewise.
8323         * tree-sra.c (scan_function): Likewise.
8324         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
8325         (optimize_stack_restore): Likewise.
8326         (pass_fold_builtins::execute): Likewise.
8327         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
8328         (mark_all_reaching_defs_necessary_1): Likewise.
8329         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
8330         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
8331         (pass_forwprop::execute): Likewise.
8332         * tree-ssa-loop-im.c (stmt_cost): Likewise.
8333         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
8334         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
8335         * tree-ssa-strlen.c (get_string_length): Likewise.
8336         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
8337         (find_func_aliases_for_call): Likewise.
8338         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
8339         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
8340         * tree-tailcall.c (find_tail_calls): Likewise.
8341         * tree.c (need_assembler_name_p): Likewise.
8342         (free_lang_data_in_decl): Likewise.
8343         (get_call_combined_fn): Likewise.
8344         * ubsan.c (is_ubsan_builtin_p): Likewise.
8345         * varasm.c (incorporeal_function_p): Likewise.
8346         * tree.h (DECL_BUILT_IN): Remove and replace with
8347         fndecl_built_in_p.
8348         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
8349         (fndecl_built_in_p): New.
8351 2018-08-27  Martin Liska  <mliska@suse.cz>
8353         PR tree-optimization/86847
8354         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
8355         Dump also subtree probability.
8356         (switch_decision_tree::do_jump_if_equal): New function.
8357         (switch_decision_tree::emit_case_nodes): Handle special
8358         situations in balanced tree that can be emitted much simpler.
8359         Fix calculation of probabilities that happen in tree expansion.
8360         * tree-switch-conversion.h (struct cluster): Add
8361         is_single_value_p.
8362         (struct simple_cluster): Likewise.
8363         (struct case_tree_node): Add new function has_child.
8364         (do_jump_if_equal): New.
8366 2018-08-27  Martin Liska  <mliska@suse.cz>
8368         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
8369         Add new argument to bit_test_cluster constructor.
8370         (bit_test_cluster::emit): Set bits really number of values
8371         handlel by a test.
8372         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
8373         probability argument.
8374         * tree-switch-conversion.h (struct bit_test_cluster):
8375         Add m_handles_entire_switch.
8377 2018-08-27  Martin Liska  <mliska@suse.cz>
8379         PR tree-optimization/86702
8380         * tree-switch-conversion.c (jump_table_cluster::emit):
8381         Make probabilities even for values in jump table
8382         according to number of cases handled.
8383         (switch_decision_tree::compute_cases_per_edge): Pass
8384         argument to reset_out_edges_aux function.
8385         (switch_decision_tree::analyze_switch_statement): Likewise.
8386         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
8387         Make it static.
8389 2018-08-27  Martin Liska  <mliska@suse.cz>
8391         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
8392         cfun argument explicitly.
8393         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
8394         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
8395         function gimple_switch_default_bb.
8396         (convert_switch_statements):
8397         (expand_builtins):
8398         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
8399         * stmt.c (label_to_block_fn): Use label_to_block and pass
8400         cfun argument explicitly and use gimple_switch_label_bb.
8401         (expand_case): Likewise.
8402         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
8403         cfun argument explicitly. Likewise.
8404         (make_edges_bb): Likewise.
8405         (make_cond_expr_edges): Likewise.
8406         (get_cases_for_edge): Likewise.
8407         (make_gimple_switch_edges): Likewise.
8408         (label_to_block_fn): Likewise.
8409         (label_to_block): Likewise.
8410         (make_goto_expr_edges): Likewise.
8411         (make_gimple_asm_edges): Likewise.
8412         (main_block_label): Likewise.
8413         (group_case_labels_stmt): Likewise.
8414         (find_taken_edge_computed_goto): Likewise.
8415         (find_taken_edge_switch_expr): Likewise.
8416         (gimple_verify_flow_info): Likewise.
8417         (gimple_redirect_edge_and_branch): Likewise.
8418         (gimple_switch_label_bb): New function.
8419         (gimple_switch_default_bb): Likewise.
8420         (gimple_switch_edge): Likewise.
8421         (gimple_switch_default_edge): Likewise.
8422         * tree-cfg.h (label_to_block_fn): Remove and replace ...
8423         (label_to_block): ... with this.
8424         (gimple_switch_label_bb): New.
8425         (gimple_switch_default_bb): Likewise.
8426         (gimple_switch_edge): Likewise.
8427         (gimple_switch_default_edge): Likewise.
8428         * tree-cfgcleanup.c (convert_single_case_switch): Use
8429         new gimple functions and pass new argument to label_to_block.
8430         (cleanup_control_flow_bb):
8431         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
8432         cfun argument explicitly.
8433         (make_eh_edges): Likewise.
8434         (redirect_eh_dispatch_edge): Likewise.
8435         (lower_resx): Likewise.
8436         (lower_eh_dispatch): Likewise.
8437         (maybe_remove_unreachable_handlers): Likewise.
8438         (unsplit_eh): Likewise.
8439         (cleanup_empty_eh): Likewise.
8440         (verify_eh_edges): Likewise.
8441         (verify_eh_dispatch_edge): Likewise.
8442         * tree-ssa-dom.c (record_edge_info): Likewise.
8443         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
8444         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
8445         (thread_through_normal_block): Likewise.
8446         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
8447         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
8448         * tree-switch-conversion.c (switch_conversion::collect): Use new
8449         gimple functions.
8450         (switch_conversion::check_final_bb): Likewise.
8451         (switch_conversion::gather_default_values): Pass new argument
8452         to label_to_block.
8453         (switch_conversion::build_constructors): Likewise.
8454         (switch_decision_tree::compute_cases_per_edge): Use new
8455         gimple_switch_edge function.
8456         (switch_decision_tree::analyze_switch_statement): Pass new argument
8457         to label_to_block.
8458         (switch_decision_tree::try_switch_expansion): Use
8459         gimple_switch_default_edge.
8460         * tree-vrp.c (find_switch_asserts): Pass new argument
8461         to label_to_block.
8462         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
8463         (vr_values::simplify_switch_using_ranges): Likewise.
8465 2018-08-27  Richard Biener  <rguenther@suse.de>
8467         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
8468         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
8470         * tree-ssa-sccvn.h (struct vn_pval): New structure.
8471         (struct vn_nary_op_s): Add unwind_to member.  Add
8472         predicated_values flag and put result into a union together
8473         with a linked list of vn_pval.
8474         (struct vn_ssa_aux): Add name member to make maintaining
8475         a map of SSA name to vn_ssa_aux possible.  Remove no longer
8476         needed info, dfsnum, low, visited, on_sccstack, use_processed
8477         and range_info_anti_range_p members.
8478         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
8479         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
8480         New functions.
8481         (vn_valueize): New global.
8482         (vn_context_bb): Likewise.
8483         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
8484         VN_INFO_PTR_INFO): Remove.
8485         * tree-ssa-sccvn.c: ... (rewrite)
8486         (pass_fre::execute): For -O2+ initialize loops and run
8487         RPO VN in optimistic mode (iterating).  For -O1 and -Og
8488         run RPO VN in non-optimistic mode.
8489         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
8490         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
8491         * doc/invoke.texi (sccvn-max-scc-size): Remove.
8492         (rpo-vn-max-loop-depth): Document.
8493         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
8494         when valuezing the VUSE signals we walked out of the region.
8495         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
8496         (phi_translate): Set VN context block to use for availability
8497         lookup.
8498         (compute_avail): Likewise.
8499         (pre_valueize): New function.
8500         (pass_pre::execute): Adjust to the RPO VN API.
8502         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
8503         (propagate_constants_for_unrolling): Remove.
8504         (tree_unroll_loops_completely): Perform value-numbering
8505         on the unrolled bodies loop parent.
8507 2018-08-27  Richard Biener  <rguenther@suse.de>
8509         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
8510         for partial antic compute.
8512 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
8514         PR rtl-optimization/87065
8515         * combine.c (simplify_if_then_else): Formatting fix.
8516         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
8517         check.
8518         (known_cond): Don't return const_true_rtx for vector modes.  Use
8519         CONST0_RTX instead of const0_rtx.  Formatting fixes.
8521 2018-08-27  Martin Liska  <mliska@suse.cz>
8523         PR gcov-profile/87069
8524         * gcov.c (process_file): Record files already processed
8525         and warn about a file being processed multiple times.
8527 2018-08-27  Martin Liska  <mliska@suse.cz>
8529         PR driver/83193
8530         * config/aarch64/aarch64.c (aarch64_override_options_internal):
8531         Set default values for x_aarch64_*_string strings.
8532         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
8533         prefix.  For -mabi do not print '=ABI' in help and use
8534         <option_value> format for -msve-vector-bits and -moverride
8535         options.
8537 2018-08-26  Jeff Law  <law@redhat.com>
8539         * config/mips/frame-header-opt.c: Include "backend.h" rather than
8540         "cfg.h"
8542 2018-08-26  Marek Polacek  <polacek@redhat.com>
8544         PR c++/87029, Implement -Wredundant-move.
8545         * doc/invoke.texi: Document -Wredundant-move.
8547 2018-08-25  Martin Sebor  <msebor@redhat.com>
8549         PR tree-optimization/87059
8550         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
8551         to the same type as the other.
8552         * fold-const.c (fold_binary_loc): Assert expectation.
8554 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
8556         * config/darwin.c (machopic_legitimize_pic_address): Clean up
8557         extraneous parentheses, dead code section and formatting.
8559 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
8561         PR c++/87091
8562         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
8563         wide enough for jumps in the line-numbering to be visible.
8564         (layout::print_gap_in_line_numbering): New member function.
8565         (layout::calculate_line_spans): When using line numbering, merge
8566         line spans that are only 1 line apart.
8567         (diagnostic_show_locus): When printing line numbers, show gaps in
8568         line numbering directly, rather than printing headers.
8569         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
8570         line-numbering with multiple line spans.
8571         (selftest::test_fixit_insert_containing_newline_2): Add test of
8572         line-numbering, in which the spans are close enough to be merged.
8574 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
8576         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
8577         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
8578         * tree-vrp.c (range_is_nonnull): Remove.
8579         (range_includes_zero_p): Accept value_range instead of min/max.
8580         (extract_range_from_binary_expr_1): Do not early bail on
8581         POINTER_PLUS_EXPR.
8582         Use range_includes_zero_p instead of range_is_nonnull.
8583         (extract_range_from_unary_expr): Use range_includes_zero_p instead
8584         of range_is_nonnull.
8585         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
8586         special case VR_ANTI_RANGE.
8587         (vrp_finalize): Same.
8588         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
8589         instead of min/max.
8590         (range_is_nonnull): Remove.
8591         * vr-values.c (vrp_stmt_computes_nonzero): Use
8592         range_includes_zero_p instead of range_is_nonnull.
8593         (extract_range_basic): Pass value_range to range_includes_zero_p
8594         instead of range_is_nonnull.
8596 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
8598         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
8599         * emit-rtl.h (rtl_data): Remove return_bnd.
8600         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
8601         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
8602         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
8603         (POINTER_BOUNDS_MODE): Remove definition.
8604         (make_pointer_bounds_mode): Remove.
8605         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
8606         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
8607         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
8608         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
8609         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
8610         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
8611         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
8613         * config/i386/i386-modes.def (BND32, BND64): Remove.
8614         * config/i386/i386.c (dbx_register_map): Remove bound registers.
8615         (dbx64_register_map): Ditto.
8616         (svr4_dbx_register_map): Ditto.
8617         (indirect_thunk_bnd_needed): Remove.
8618         (indirect_thunks_bnd_used): Ditto.
8619         (indirect_return_bnd_needed): Ditto.
8620         (indirect_return_via_cx_bnd): Ditto.
8621         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
8622         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
8623         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
8624         (output_indirect_thunk_function): Remove handling of
8625         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
8626         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
8627         (ix86_save_reg): Remove handling of crtl->return_bnd.
8628         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
8629         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
8630         and UNSPEC_BNDLX_ADDR.
8631         (ix86_output_indirect_branch_via_reg): Remove handling of
8632         indirect_thunk_prefix_bnd.
8633         (ix86_output_indirect_branch_via_push): Ditto.
8634         (ix86_output_function_return): Ditto.
8635         (ix86_output_indirect_function_return): Ditto.
8636         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
8637         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
8638         (CALL_USED_REGISTERS): Ditto.
8639         (REG_ALLOC_ORDER): Update for removal of bound registers.
8640         (HI_REGISTER_NAMES): Ditto.
8641         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
8642         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
8643         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
8644         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
8645         (FIRST_PSEUDO_REG): Update.
8646         (BND): Remove mode iterator.
8647         * config/i386/predicates.md (bnd_mem_operator): Remove.
8649 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
8651         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
8652         vectors.
8654 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
8656         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
8657         the case in which the permute needs only a single element and
8658         repeats for every vector of the result.  Extend that case to
8659         handle variable-length vectors.
8660         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
8662 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
8664         PR debug/79342
8665         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
8666         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
8668 2018-08-24  Richard Biener  <rguenther@suse.de>
8670         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
8671         bb_flags_allocated members.
8672         (auto_flag): New RAII class for allocating flags.
8673         (auto_edge_flag): New RAII class for allocating edge flags.
8674         (auto_bb_flag): New RAII class for allocating bb flags.
8675         * cfgloop.c (verify_loop_structure): Allocate temporary edge
8676         flag dynamically.
8677         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
8678         in favor of temporarily allocated BB flag.
8679         * hsa-brig.c: Re-order includes.
8680         * hsa-dump.c: Likewise.
8681         * hsa-regalloc.c: Likewise.
8682         * print-rtl.c: Likewise.
8683         * profile-count.c: Likewise.
8685 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
8687         PR target/86989
8688         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
8689         the TOC register.
8691 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
8693         PR 87073/bootstrap
8694         * wide-int-range.cc (wide_int_range_div): Do not ignore result
8695         from wide_int_range_multiplicative_op.
8697 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
8699         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
8700         "permutaion".
8702 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
8704         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
8705         to 'expanded'.
8707 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
8709         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
8710         full GENERIC expression used for replacement.
8712 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
8714         * tree-vrp.c (abs_extent_range): Remove.
8715         (extract_range_into_wide_ints): Pass wide ints by reference.
8716         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
8717         Pass wide ints by reference in all calls to
8718         extract_range_into_wide_ints.
8719         * wide-int-range.cc (wide_int_range_div): New.
8720         * wide-int-range.h (wide_int_range_div): New.
8721         (wide_int_range_includes_zero_p): New.
8722         (wide_int_range_zero_p): New.
8724 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
8726         * config/aarch64/aarch64.md (arches): New enum.
8727         (arch): New enum attr.
8728         (arch_enabled): New attr.
8729         (enabled): Now uses arch_enabled only.
8730         (simd, sve, fp16): Removed attribute.
8731         (fp): Attr now defined in terms of 'arch'.
8732         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
8733         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
8734         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
8735         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
8736         attributes into 'arch'.
8737         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
8738         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
8739         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
8740         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
8741         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
8742         'simd' attribute into 'arch'.
8743         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
8744         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
8745         Convert use of 'fp' attribute to 'arch'.
8746         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
8747         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
8748         into 'arch'.
8749         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
8750         (different modes) Merge 'fp' and 'simd' into 'arch'.
8751         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
8752         'simd' into 'arch'.
8754 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
8756         PR rtl-optimization/87026
8757         * expmed.c (canonicalize_comparison): If we can no longer create
8758         pseudoregisters, don't.
8760 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
8762         PR target/86951
8763         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
8764         prototype.
8765         * config/arm/arm.c (speculation_barrier_libfunc): New static
8766         variable.
8767         (arm_init_libfuncs): Initialize it.
8768         (arm_emit_speculation_barrier): New function.
8769         * config/arm/arm.md (speculation_barrier): Call
8770         arm_emit_speculation_barrier for architectures that do not have
8771         DSB or ISB.
8772         (speculation_barrier_insn): Only match on Armv7 or later.
8774 2018-08-23  Richard Biener  <rguenther@suse.de>
8776         PR middle-end/87024
8777         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
8778         calls.
8780 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
8782         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
8783         of single-vector TBLs.
8784         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
8785         one input is given.
8787 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
8789         PR target/85910
8790         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
8791         aarch64_evpc_tbl guard.
8793 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8795         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
8796         behaviour.
8798 2018-08-22  Martin Sebor  <msebor@redhat.com>
8800         PR middle-end/87052
8801         * tree-pretty-print.c (pretty_print_string): Add argument.
8802         (dump_generic_node): Call to pretty_print_string with string size.
8804 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
8806         PR rtl-optimization/86771
8807         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
8808         of two SETs into those two SETs, one to be placed at i2, if that SETs
8809         destination is modified between i2 and i3.
8811 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8813         PR tree-optimization/86725
8814         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
8815         function.
8816         (vect_analyze_scalar_cycles_1): Check it.
8818 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8820         PR tree-optimization/86725
8821         * tree-vect-loop.c (vect_is_simple_reduction): When treating
8822         an outer loop phi as a double reduction, make sure that the
8823         single user of the phi result is an inner loop phi.
8825 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8827         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
8828         grouped stores with gaps to a strided group.
8830 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8832         * tree-vect-stmts.c (get_group_load_store_type)
8833         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
8834         first statement in a group.
8836 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8838         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
8839         the sequence used in gcc/gcc.c.
8841 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8843         PR other/704
8844         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
8845         building it.
8847 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8849         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
8850         Darwin10-specific unwinder-shim.
8851         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
8852         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
8853         New to cater for Darwin10 Rosetta.
8855 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8857         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
8858         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
8860 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8862         PR bootstrap/81033
8863         PR target/81733
8864         PR target/52795
8865         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
8866         (dwarf2out_switch_text_section): Generate a local label for the
8867         second function sub-section and apply it as the second FDE start
8868         label.
8869         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
8870         second sub-section start.
8872 2018-08-22  Richard Biener  <rguenther@suse.de>
8874         PR tree-optimization/86988
8875         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
8877 2018-08-22  Richard Biener  <rguenther@suse.de>
8879         PR tree-optimization/86945
8880         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
8882 2018-08-22  Alexandre Oliva <oliva@adacore.com>
8884         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
8885         a comment about how uses of r2 for .sdata2 come about.
8887 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
8889         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
8891 2018-08-21  Marek Polacek  <polacek@redhat.com>
8893         PR c++/86981, Implement -Wpessimizing-move.
8894         * doc/invoke.texi: Document -Wpessimizing-move.
8896 2018-08-21  Jan Hubicka  <jh@suse.cz>
8898         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
8899         * tree.h (is_redundant_typedef): Remove.
8900         * dwarf2out.c (is_redundant_typedef): Turn into static function.
8902 2018-08-21  Jan Hubicka  <jh@suse.cz>
8904         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
8905         when possible.
8907 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
8909         * expmed.c (extract_low_bits): Reject invalid subregs early.
8911 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8913         PR middle-end/86121
8914         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
8915         behaviour.
8917 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8919         * config/vxworks.h: Guard vxworks_asm_out_constructor and
8920         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
8921         * config/vxworks.c: Likewise.
8923 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8925         * config/vxworks.c: Set targetm.have_ctors_dtors
8926         if HAVE_INITFINI_ARRAY_SUPPORT.
8927         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
8928         if HAVE_INITFINI_ARRAY_SUPPORT.
8930 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8932         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
8933         default search path for VxWorks < 7.
8935 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8937         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
8938         (get_format_string): Refer to c_getstr.
8940 2018-08-21  Tom de Vries  <tdevries@suse.de>
8942         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
8943         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
8944         (debuginfo_early_stop): Declare.
8945         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
8946         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
8947         (debuginfo_early_stop): New function.
8948         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
8949         and debuginfo_early_stop.
8950         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
8951         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
8952         (general_init): Call debuginfo_early_init.
8953         (finalize): Call debuginfo_fini.
8954         (do_compile): Call debuginfo_init.
8955         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
8956         -fdump-early-debug.
8957         (@item -fdump-debug, @item -fdump-earlydebug): Add.
8959 2018-08-21  Tom de Vries  <tdevries@suse.de>
8961         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
8962         flag_dump_noaddr and flag_dump_unnumbered.
8964 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
8966         * wide-int-range.cc (wide_int_range_abs): New.
8967         (wide_int_range_order_set): Rename from wide_int_range_min_max.
8968         * wide-int-range.h (wide_int_range_abs): New.
8969         (wide_int_range_min_max): New.
8970         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
8971         case to call wide_int_range_abs.
8972         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
8973         (extract_range_from_abs_expr): Delete.
8975 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
8977         PR target/87033
8978         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
8979         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
8980         for indexed loads.
8982 2018-08-20  Nathan Sidwell  <nathan@acm.org>
8983             Jeff Law <law@redhat.com>
8985         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
8986         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
8988 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
8990         PR other/84889
8991         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
8992         (decl_attributes): Likewise.
8993         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
8994         instance.
8995         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
8996         * diagnostic-core.h (class auto_diagnostic_group): New class.
8997         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
8998         (diagnostic_report_diagnostic): Handle the first diagnostics within
8999         a group.
9000         (emit_diagnostic): Add auto_diagnostic_group instance.
9001         (inform): Likewise.
9002         (inform_n): Likewise.
9003         (warning): Likewise.
9004         (warning_at): Likewise.
9005         (warning_n): Likewise.
9006         (pedwarn): Likewise.
9007         (permerror): Likewise.
9008         (error): Likewise.
9009         (error_n): Likewise.
9010         (error_at): Likewise.
9011         (sorry): Likewise.
9012         (fatal_error): Likewise.
9013         (internal_error): Likewise.
9014         (internal_error_no_backtrace): Likewise.
9015         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
9016         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
9017         * diagnostic.h (struct diagnostic_context): Add fields
9018         "diagnostic_group_nesting_depth",
9019         "diagnostic_group_emission_count", "begin_group_cb",
9020         "end_group_cb".
9021         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
9022         Add auto_diagnostic_group instance(s).
9023         (find_explicit_erroneous_behavior): Likewise.
9024         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
9025         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
9026         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
9027         (gimplify_va_arg_expr): Likewise.
9028         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
9029         (HSA_SORRY_AT): Likewise.
9030         * ipa-devirt.c (compare_virtual_tables): Likewise.
9031         (warn_odr): Likewise.
9032         * multiple_target.c (expand_target_clones): Likewise.
9033         * opts-common.c (cmdline_handle_error): Likewise.
9034         * reginfo.c (globalize_reg): Likewise.
9035         * substring-locations.c (format_warning_n_va): Likewise.
9036         * tree-inline.c (expand_call_inline): Likewise.
9037         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
9038         * tree-ssa-loop-niter.c
9039         (do_warn_aggressive_loop_optimizations): Likewise.
9040         * tree-ssa-uninit.c (warn_uninit): Likewise.
9041         * tree.c (warn_deprecated_use): Likewise.
9043 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
9045         PR target/87014
9046         * config/i386/i386.md (eh_return): Always update EH return
9047         address in word_mode.
9049 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
9051         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
9052         TARGET_SPLIT_COMPLEX_ARG is defined.
9054 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9056         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
9058 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9060         PR target/86984
9061         * expr.c (expand_assignment): Assert that bitpos is positive.
9062         (store_field): Likewise
9063         (expand_expr_real_1): Make sure that bitpos is positive.
9064         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
9065         integer overflow.
9067 2018-08-20  Nathan Sidwell  <nathan@acm.org>
9069         * Makefile.in (CPP_ID_DATA_H): Delete.
9070         (CPP_INTERNAL_H): Don't add it.
9071         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
9072         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
9074 2018-08-20  Richard Biener  <rguenther@suse.de>
9076         PR tree-optimization/78655
9077         * tree-vrp.c (extract_range_from_binary_expr_1): Make
9078         pointer + offset nonnull if either operand is nonnull work.
9080 2018-08-20  Tom de Vries  <tdevries@suse.de>
9082         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
9083         unless the referenced die describes the added property using
9084         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
9085         Otherwise, add a DW_AT_location to the referenced die.
9087 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
9089         PR target/86994
9090         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
9091         register_operand when calling ix86_set_reg_reg_cost.
9092         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
9093         Set *total to 0 for operands that satisfy x86_64_immediate_operand
9094         predicate and to 1 otherwise.
9096 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
9098         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
9099         emit a diagnostic that it is not supported and reset the option.
9100         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
9101         supported and consume the option.  (ASM_FINAL_SPEC): New.
9103 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
9105         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
9106         a sentence.
9108 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
9110         C-SKY port: Documentation
9112         * doc/extend.texi (C-SKY Function Attributes): New section.
9113         * doc/invoke.texi (Option Summary): Add C-SKY options.
9114         (C-SKY Options): New section.
9115         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
9117 2018-08-17  Jojo  <jijie_rong@c-sky.com>
9118             Huibin Wang  <huibin_wang@c-sky.com>
9119             Sandra Loosemore  <sandra@codesourcery.com>
9120             Chung-Lin Tang  <cltang@codesourcery.com>
9122         C-SKY port: Backend implementation
9124         * config/csky/*: New.
9125         * common/config/csky/*: New.
9127 2018-08-17  Jojo  <jijie_rong@c-sky.com>
9128             Huibin Wang  <huibin_wang@c-sky.com>
9129             Sandra Loosemore  <sandra@codesourcery.com>
9130             Chung-Lin Tang  <cltang@codesourcery.com>
9131             Andrew Jenner  <andrew@codesourcery.com>
9133         C-SKY port: Configury
9135         * config.gcc (csky-*-*): New.
9136         * configure.ac: Add csky to targets for dwarf2 debug_line support.
9137         * configure: Regenerated.
9139 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
9141         * dump-context.h: Include "dumpfile.h".
9142         (dump_context::dump_printf_va): Convert final param from va_list
9143         to va_list *.  Convert from ATTRIBUTE_PRINTF to
9144         ATTRIBUTE_GCC_DUMP_PRINTF.
9145         (dump_context::dump_printf_loc_va): Likewise.
9146         * dumpfile.c: Include "stringpool.h".
9147         (make_item_for_dump_printf_va): Delete.
9148         (make_item_for_dump_printf): Delete.
9149         (class dump_pretty_printer): New class.
9150         (dump_pretty_printer::dump_pretty_printer): New ctor.
9151         (dump_pretty_printer::emit_items): New member function.
9152         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
9153         function.
9154         (dump_pretty_printer::emit_item): New member function.
9155         (dump_pretty_printer::stash_item): New member function.
9156         (dump_pretty_printer::format_decoder_cb): New member function.
9157         (dump_pretty_printer::decode_format): New member function.
9158         (dump_context::dump_printf_va): Reimplement in terms of
9159         dump_pretty_printer.
9160         (dump_context::dump_printf_loc_va): Convert final param from va_list
9161         to va_list *.
9162         (dump_context::begin_scope): Reimplement call to
9163         make_item_for_dump_printf.
9164         (dump_printf): Update for change to dump_printf_va.
9165         (dump_printf_loc): Likewise.
9166         (selftest::test_capture_of_dump_calls): Convert "stmt" from
9167         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
9168         with %T, %E, and %G.
9169         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
9170         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
9171         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
9172         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
9173         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
9174         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
9175         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
9176         within a dump_printf_loc call to "%wu".
9177         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
9178         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
9179         missing space after "=".
9180         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
9181         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
9182         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
9183         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
9184         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
9185         duplicate "vectorized" from message.
9187 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
9189         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
9190         polyNxK_t element's TYPE_STRING_FLAG.
9192 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
9194         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
9195         (they were unnamed before).  Fix comments.
9197 2018-08-17  Nathan Sidwell  <nathan@acm.org>
9199         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
9201 2018-08-17  Richard Biener  <rguenther@suse.de>
9203         PR tree-optimization/86841
9204         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
9206 2018-08-17  Martin Liska  <mliska@suse.cz>
9208         * common.opt: Remove Warn, Init and Report for options with
9209         Ignore/Deprecated flag. Warning is done automatically for
9210         Deprecated flags.
9211         * config/i386/i386.opt: Likewise.
9212         * config/ia64/ia64.opt: Likewise.
9213         * config/rs6000/rs6000.opt: Likewise.
9214         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
9215         Remove usage of flag_check_pointer_bounds.
9216         * lto-wrapper.c (merge_and_complain): Do not handle
9217         OPT_fcheck_pointer_bounds.
9218         (append_compiler_options): Likewise.
9219         * opt-functions.awk: Do not handle Deprecated.
9220         * optc-gen.awk: Check that Var, Report and Init are not
9221         used for an option with Ignore/Deprecated flag.
9222         * opts-common.c (decode_cmdline_option): Do not report
9223         CL_ERR_DEPRECATED.
9224         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
9225         options.
9226         * opts.h (struct cl_option): Remove cl_deprecated flag.
9227         (CL_ERR_DEPRECATED): Remove error enum value.
9229 2018-08-17  Richard Biener  <rguenther@suse.de>
9231         PR middle-end/86505
9232         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
9233         across a va-arg-pack using call adjust its return value accordingly.
9235 2018-08-16  Martin Sebor  <msebor@redhat.com>
9237         PR tree-optimization/86853
9238         * gimple-ssa-sprintf.c (struct format_result): Rename member.
9239         (struct fmtresult): Add member and initialize it in ctors.
9240         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
9241         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
9242         (format_directive): Set POSUNDER4K when MAYFAIL is set.
9243         (parse_directive): Handle %C same as %c.
9244         (sprintf_dom_walker::compute_format_length): Adjust.
9245         (is_call_safe): Adjust.
9247 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9249         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
9250         for determining how to count the elements.
9251         * builtins.h (c_strlen): Adjust prototype.
9252         * expr.c (string_constant): Add new parameter mem_size.
9253         Set *mem_size appropriately.
9254         * expr.h (string_constant): Adjust protoype.
9255         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
9256         * gimple-fold.h (get_range_strlen): Adjust prototype.
9257         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
9258         (format_string): Call get_string_length with eltsize.
9260 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
9262         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
9263         to emit the span, rather than setting it as the prefix.
9265 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
9267         * diagnostic-show-locus.c (layout::start_annotation_line): Add
9268         "margin_char" parameter, defaulting to space.  Use it in place
9269         of pp_space for the initial part of the margin.
9270         (layout::print_leading_fixits): Use '+' when filling the margin
9271         of line-insertion fix-it hints.
9273 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9275         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
9276         Delete.
9278 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9280         * config/rs6000/altivec.md: Don't set length attribute to the default
9281         value.
9282         * config/rs6000/darwin.md: Ditto.
9283         * config/rs6000/dfp.md: Ditto.
9284         * config/rs6000/htm.md: Ditto.
9285         * config/rs6000/rs6000.md: Ditto.
9286         * config/rs6000/sync.md: Ditto.
9287         * config/rs6000/vsx.md: Ditto.
9289 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9291         * config/rs6000/altivec.md: Don't set length attribute to the default
9292         value, for branch instructions.
9293         * config/rs6000/darwin.md: Ditto.
9294         * config/rs6000/rs6000.md: Ditto.
9296 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9298         * config/rs6000/rs6000.md (length): Always define as const_int 4.
9299         (unnamed conditional branch define_insn): Set length to 4 or 8
9300         depending on offset.
9301         (<bd>_<mode>): Similar, for alternative 0.
9302         (<bd>tf_<mode>): Ditto.
9304 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
9306         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
9308 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
9310         * doc/rtl.texi: Replace old RTX class names with new names.
9313 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
9315         * expmed.h (canonicalize_comparison): New declaration.
9316         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
9317         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
9318         * optabs.c (prepare_cmp_insn): Likewise.
9319         * rtl.h (unsigned_condition_p): New function which checks if a
9320         comparison operator is unsigned.
9322 2018-08-16  Nathan Sidwell  <nathan@acm.org>
9324         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
9325         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
9327 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
9329         PR target/84711
9330         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
9331         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
9332         (mov<mov>): ..this and enable unconditionally.
9334 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
9336         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
9338 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
9340         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
9341         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
9342         "Common" with "Target".
9344 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
9346         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
9347         * doc/invoke.texi (mmitigate-rop): Remove.
9348         * config/i386/i386.c: Do not include "regrename.h".
9349         (ix86_rop_should_change_byte_p, reg_encoded_number)
9350         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
9351         Remove.
9352         (ix86_reorg): Remove call to ix86_mitigate_rop.
9353         * config/i386/i386.md (attr "modrm_class"): Remove.
9354         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
9355         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
9356         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
9358 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
9360         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
9361         allow folding of mergeh() and mergel() for the float and double types.
9362         (fold_mergehl_helper): Rework to handle building a permute tree
9363         for float vectors.
9365 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
9367         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
9368         for TARGET_SSE.
9370 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
9372         * common.opt (fdiagnostics-show-labels): New option.
9373         * diagnostic-show-locus.c (class layout_range): Add field
9374         "m_label".
9375         (class layout): Add field "m_show_labels_p".
9376         (layout_range::layout_range): Add param "label" and use it to
9377         initialize m_label.
9378         (make_range): Pass in NULL for new "label" param of layout_range's
9379         ctor.
9380         (layout::layout): Initialize m_show_labels_p.
9381         (layout::maybe_add_location_range): Pass in loc_range->m_label
9382         when constructing layout_range instances.
9383         (struct line_label): New struct.
9384         (layout::print_any_labels): New member function.
9385         (layout::print_line): Call it if label-printing is enabled.
9386         (selftest::test_one_liner_labels): New test.
9387         (selftest::test_diagnostic_show_locus_one_liner): Call it.
9388         * diagnostic.c (diagnostic_initialize): Initialize
9389         context->show_labels_p.
9390         * diagnostic.h (struct diagnostic_context): Add field
9391         "show_labels_p".
9392         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
9393         -fno-diagnostics-show-labels.
9394         * dwarf2out.c (gen_producer_string): Add
9395         OPT_fdiagnostics_show_labels to the ignored options.
9396         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
9397         param.
9398         (gcc_rich_location::maybe_add_expr): Likewise.
9399         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
9400         label" param, defaulting to NULL.
9401         (gcc_rich_location::add_expr): Add "label" param.
9402         (gcc_rich_location::maybe_add_expr): Likewise.
9403         (class text_range_label): New class.
9404         (class range_label_for_type_mismatch): New class.
9405         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
9406         of format_warning_va.
9407         (fmtwarn_n): Likewise for new params of format_warning_n_va.
9408         * lto-wrapper.c (merge_and_complain): Add
9409         OPT_fdiagnostics_show_labels to the "pick one setting" options.
9410         (append_compiler_options): Likewise to the dropped options.
9411         (append_diag_options): Likewise to the passed-on options.
9412         * opts.c (common_handle_option): Handle the new option.
9413         * selftest-diagnostic.c
9414         (test_diagnostic_context::test_diagnostic_context): Enable
9415         show_labels_p.
9416         * substring-locations.c: Include "gcc-rich-location.h".
9417         (format_warning_n_va): Add "fmt_label" and "param_label" params
9418         and use them as appropriate.
9419         (format_warning_va): Add "fmt_label" and "param_label" params,
9420         passing them on to format_warning_n_va.
9421         (format_warning_at_substring): Likewise.
9422         (format_warning_at_substring_n): Likewise.
9423         * substring-locations.h (format_warning_va): Add "fmt_label" and
9424         "param_label" params.
9425         (format_warning_n_va): Likewise.
9426         (format_warning_at_substring): Likewise.
9427         (format_warning_at_substring_n): Likewise.
9428         * toplev.c (general_init): Initialize global_dc->show_labels_p.
9430 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
9432         PR testsuite/86519
9433         * builtins.c (expand_builtin_memcmp): Do not expand the call
9434         when overflow is detected.
9436 2018-08-15  Martin Sebor  <msebor@redhat.com>
9438         PR tree-optimization/71625
9439         * config/aarch64/aarch64-builtins.c
9440         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
9442 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
9444         * config/s390/s390.c (s390_reorg): Remove loop.
9446 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
9448         * config/darwin.c
9449          (darwin_function_switched_text_sections): Delete.
9450         * gcc/config/darwin.h
9451          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
9453 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
9455         PR target/81685
9456         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
9457         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
9458         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
9460 2018-08-15  Martin Liska  <mliska@suse.cz>
9462         PR tree-optimization/86925
9463         * predict.c (expr_expected_value_1): When taking
9464         later predictor, assign also probability.
9465         Use fold_build2_initializer_loc in order to fold
9466         the expression in -frounding-math.
9468 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
9470         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
9471         patterns.
9472         (expand_vec_perm_1): Try the new method.
9474 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
9476         PR target/86547
9477         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
9478         Check whether lra_live_max_point is 0 before dividing.
9480 2018-08-14  Martin Sebor  <msebor@redhat.com>
9482         PR tree-optimization/86650
9483         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
9484         (vrp_prop::check_mem_ref): Same.
9486 2018-08-13  Liu Hao <lh_mouse@126.com>
9488         * pretty-print.c (eat_esc_sequence): Swap the foreground and
9489         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
9490         and clear it thereafter, as it only works for DBCS.
9492 2018-08-13  Liu Hao <lh_mouse@126.com>
9494         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
9495         handle returned by _get_osf_handle().
9497 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
9499         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
9500         for folding vec_perm.
9502 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
9504         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
9505         Add support for gimple-folding of vec_pack() and vec_unpack()
9506         intrinsics.
9508 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
9510         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
9511         vec_xst variants to the list.
9512         (rs6000_gimple_fold_builtin): Add support for folding unaligned
9513         vector loads and stores.
9515 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
9517         * config.gcc (rs6000-ibm-aix4.x): Delete.
9518         (rs6000-ibm-aix5.1): Delete.
9519         (rs6000-ibm-aix5.2): Delete.
9520         (rs6000-ibm-aix5.3): Delete.
9521         * config/rs6000/aix43.h: Delete.
9522         * config/rs6000/aix51.h: Delete.
9523         * config/rs6000/aix52.h: Delete.
9524         * config/rs6000/t-aix43: Delete.
9526 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
9528         * config/s390/s390.c (s390_decompose_constant_pool_ref):
9529         New function.
9530         (s390_decompose_address): Factor out constant pool ref
9531         decomposition.
9533 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9535         * config/nds32/nds32-predicates.c
9536         (nds32_can_use_bclr_p): Change return type as bool.
9537         (nds32_can_use_bset_p): Ditto.
9538         (nds32_can_use_btgl_p): Ditto.
9539         (nds32_can_use_bitci_p): Ditto.
9540         * config/nds32/nds32-protos.h
9541         (nds32_can_use_bclr_p): Change declaration.
9542         (nds32_can_use_bset_p): Ditto.
9543         (nds32_can_use_btgl_p): Ditto.
9544         (nds32_can_use_bitci_p): Ditto.
9546 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9548         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
9549         Support -msched-prolog-epilog option.
9550         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
9552 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9554         * common/config/nds32/nds32-common.c
9555         (nds32_option_optimization_table): Enalbe -malways-align.
9557 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9559         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
9560         extra_headers.
9561         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
9562         OPT_misr_secure_ case.
9563         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
9564         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
9565         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
9566         secure attribute.
9567         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
9568         (nds32_isr_info): New field security_level.
9569         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
9570         * config/nds32/nds32.md (return_internal): Consider critical attribute.
9571         * config/nds32/nds32.opt (misr-secure): New option.
9572         * config/nds32/nds32_init.inc: New file.
9573         * config/nds32/nds32_isr.h: New file.
9575 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
9577         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
9578         Update comment for atomic instructions.
9579         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
9580         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
9581         Remove.
9582         (atomic_loaddi): Revise fence expansion to only emit fence prior to
9583         load for __ATOMIC_SEQ_CST model.
9584         (atomic_loaddi_1): Remove float register target.
9585         (atomic_storedi): Handle CONST_INT values.
9586         (atomic_storedi_1): Remove float register source.  Add special case
9587         for zero value.
9588         (memory_barrier): New expander and insn.
9590 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
9592         PR tree-optimization/86835
9593         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
9594         new_stmt after def_gsi, make sure to insert new_square_stmt after
9595         that stmt, not 2 stmts before it.
9597 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
9599         PR target/82418
9600         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
9601         instead of SWI48.
9603 2018-08-10  Martin Liska  <mliska@suse.cz>
9605         PR target/83610
9606         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
9607         function type.
9608         * builtins.c (expand_builtin_expect_with_probability):
9609         New function.
9610         (expand_builtin_expect_with_probability): New function.
9611         (build_builtin_expect_predicate): Add new argumnet probability
9612         for BUILT_IN_EXPECT_WITH_PROBABILITY.
9613         (fold_builtin_expect):
9614         (fold_builtin_2):
9615         (fold_builtin_3):
9616         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
9617         * builtins.h (fold_builtin_expect): Set new argument.
9618         * doc/extend.texi: Document __builtin_expect_with_probability.
9619         * doc/invoke.texi: Likewise.
9620         * gimple-fold.c (gimple_fold_call): Pass new argument.
9621         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
9622         also BUILT_IN_EXPECT_WITH_PROBABILITY.
9623         * predict.c (get_predictor_value): New function.
9624         (expr_expected_value): Add new argument probability. Assume
9625         that predictor and probability are always non-null.
9626         (expr_expected_value_1): Likewise.  For __builtin_expect and
9627         __builtin_expect_with_probability set probability.  Handle
9628         combination in binary expressions.
9629         (tree_predict_by_opcode): Simplify code by simply calling
9630         get_predictor_value.
9631         (pass_strip_predict_hints::execute): Add handling of
9632         BUILT_IN_EXPECT_WITH_PROBABILITY.
9633         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
9634         new predictor.
9635         * tree.h (DECL_BUILT_IN_P): New function.
9637 2018-08-10  Martin Liska  <mliska@suse.cz>
9639         PR tree-optimization/85799
9640         * passes.def: Add argument for pass_strip_predict_hints.
9641         * predict.c (class pass_strip_predict_hints): Add new argument
9642         early_p.
9643         (strip_predictor_early): New function.
9644         (pass_strip_predict_hints::execute): Call the function to
9645         strip predictors.
9646         (strip_predict_hints): New function.
9647         * predict.def: Fix comment.
9649 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
9651         * Makefile.in: Clarify which tm.texi to copy over to assert the
9652         right to grant a GFDL license for all.
9654 2018-08-09  Jeff Law  <law@redhat.com>
9656         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
9657         unused variable.
9659 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
9661         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
9662         prototype.
9664 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9666         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
9667         reductions for variable-length vectors.
9669 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
9671         PR other/84889
9672         * common.opt (fdiagnostics-show-line-numbers): New option.
9673         * diagnostic-show-locus.c (class layout): Add fields
9674         "m_show_line_numbers_p" and "m_linenum_width";
9675         (num_digits): New function.
9676         (test_num_digits): New function.
9677         (layout::layout): Initialize new fields.  Update m_x_offset
9678         logic to handle any left margin.
9679         (layout::print_source_line): Print line number when requested.
9680         (layout::start_annotation_line): New member function.
9681         (layout::print_annotation_line): Call it.
9682         (layout::print_leading_fixits): Likewise.
9683         (layout::print_trailing_fixits): Likewise.  Update calls to
9684         move_to_column for new parameter.
9685         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
9686         it to potentially call start_annotation_line.
9687         (layout::show_ruler): Call start_annotation_line.
9688         (selftest::test_line_numbers_multiline_range): New selftest.
9689         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
9690         and selftest::test_line_numbers_multiline_range.
9691         * diagnostic.c (diagnostic_initialize): Initialize
9692         show_line_numbers_p.
9693         * diagnostic.h (struct diagnostic_context): Add field
9694         "show_line_numbers_p".
9695         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
9696         -fno-diagnostics-show-line-numbers.
9697         * dwarf2out.c (gen_producer_string): Add
9698         OPT_fdiagnostics_show_line_numbers to the ignored options.
9699         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
9700         one setting" options.
9701         (append_compiler_options): Likewise to the dropped options.
9702         (append_diag_options): Likewise to the passed-on options.
9703         * opts.c (common_handle_option): Handle the new option.
9704         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
9706 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9708         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
9709         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
9710         third argument of type "const signed char" to descriptions of
9711         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
9712         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
9713         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
9714         __builtin_bcdsub_ov functions.
9716 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9718         PR tree-optimization/86858
9719         * tree-vect-loop.c (vect_is_simple_reduction): Restore
9720         flow_bb_inside_loop_p calls.
9722 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9724         PR tree-optimization/86871
9725         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
9726         instead of gimple_assign_lhs.
9728 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
9730         PR target/86887
9731         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
9732         register constraint to operand 0.
9733         (add<mode>3_carryinC): Likewise.
9734         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
9736 2018-08-09  Martin Liska  <mliska@suse.cz>
9738         PR c/86895
9739         * common.opt: Remove extra line.
9741 2018-08-09  Martin Liska  <mliska@suse.cz>
9743         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
9744         at the end of a line, make first letter capital and end up
9745         a sentence with a dot.
9746         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
9747         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
9748         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
9749         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
9750         (PARAM_MAX_ISL_OPERATIONS): Likewise.
9751         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
9752         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
9753         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
9754         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
9755         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
9756         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
9757         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
9758         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
9759         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
9760         (PARAM_TREE_REASSOC_WIDTH): Likewise.
9761         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
9762         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
9763         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
9765 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
9767         PR target/84332
9768         * config/s390/s390.c (s390_option_override_internal): Reduce the
9769         stack-clash-protection-probe-interval param if it would be too big
9770         for z900.
9772 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
9774         PR target/46179
9775         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
9776         * config/m68k/m68k.c (handle_move_double): Don't call
9777         m68k_final_prescan_insn.
9778         (m68k_adjust_decorated_operand): Renamed from
9779         m68k_final_prescan_insn, remove first and third operand and
9780         simplify.
9781         (print_operand): Call it.
9782         (print_operand_address): Call it.
9784 2018-08-08  Nathan Sidwell  <nathan@acm.org>
9786         * diagnostic.c (diagnostic_report_current_module): Use
9787         linemap_included_from & linemap_included_from_linemap.
9789 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
9791         * config/aarch64/aarch64-cores.def: Add phecda core.
9792         * config/aarch64/aarch64-tune.md: Regenerate.
9793         * doc/invoke.texi: Add phecda core.
9795 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
9797         PR target/85295
9798         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
9799         definitions.
9800         * config/s390/s390.md ("movti"): Add more alternatives for
9801         constant to GPR copies.
9803 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
9805         * config/s390/s390.c: Fix whitespace damage throughout the file.
9806         * config/s390/s390.h: Likewise.
9807         * config/s390/tpf.h: Likewise.
9809 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
9811         * config/s390/s390.c (s390_loadrelative_operand_p):
9812         Remove TARGET_CPU_ZARCH usages.
9813         (s390_rtx_costs): Likewise.
9814         (s390_legitimate_constant_p): Likewise.
9815         (s390_cannot_force_const_mem): Likewise.
9816         (legitimate_reload_constant_p): Likewise.
9817         (s390_preferred_reload_class): Likewise.
9818         (legitimize_pic_address): Likewise.
9819         (legitimize_tls_address): Likewise.
9820         (s390_split_branches): Removed.
9821         (s390_add_execute): Removed.
9822         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
9823         (s390_mainpool_start): Likewise.
9824         (s390_mainpool_finish): Likewise.
9825         (s390_mainpool_cancel): Removed.
9826         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
9827         (s390_chunkify_cancel): Likewise.
9828         (s390_return_addr_rtx): Likewise.
9829         (s390_register_info): Remove split_branches_pending_p uages.
9830         (s390_optimize_register_info): Likewise.
9831         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
9832         split_branches_pending_p usages.
9833         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
9834         (s390_load_got): Likewise.
9835         (s390_expand_split_stack_prologue): Likewise.
9836         (output_asm_nops): Likewise.
9837         (s390_function_profiler): Likewise.
9838         (s390_emit_call): Likewise.
9839         (s390_conditional_register_usage): Likewise.
9840         (s390_optimize_prologue): Likewise.
9841         (s390_reorg): Remove TARGET_CPU_ZARCH and
9842         split_branches_pending_p usages.
9843         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
9844         usages.
9845         (s390_output_indirect_thunk_function): Likewise.
9846         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
9847         (TARGET_CPU_ZARCH_P): Removed.
9848         (struct machine_function): Remove split_branches_pending_p.
9849         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
9851 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
9853         * common/config/s390/s390-common.c (processor_flags_table):
9854         Remove flags.
9855         * config.gcc: Remove with_arch/with_tune support.
9856         * config/s390/2064.md: Remove cpu attribute comparisons.
9857         * config/s390/driver-native.c (s390_host_detect_local_cpu):
9858         Remove MTN.
9859         * config/s390/linux.h (ASM_SPEC):
9860         Remove -march support.
9861         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
9862         Use a table to get an arch level.
9863         * config/s390/s390-opts.h (enum processor_type):
9864         Remove enum values.
9865         * config/s390/s390.c
9866         (processor_table): Remove entries, add arch_level values.
9867         (s390_issue_rate): Remove cases.
9868         (s390_option_override): Adjust
9869         s390_option_override_internal() call.
9870         (s390_option_override_internal): Remove deprecation warning.
9871         (s390_valid_target_attribute_tree): Adjust
9872         s390_option_override_internal() call.
9873         * config/s390/s390.h (struct s390_processor):
9874         Share with s390-c.c, add arch_level field.
9875         * config/s390/s390.md:
9876         Remove occurrences in cpu attribute.
9877         * config/s390/s390.opt: Remove -march/-mtune support.
9878         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
9879         * doc/invoke.texi: Remove deprecation warning.
9881 2018-08-08  Luis Machado  <luis.machado@linaro.org>
9883         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
9884         global.
9885         (qdf24xx_tunings): Set vector cost structure to
9886         qdf24xx_vector_cost.
9888         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
9889         <register_sextend>: Set to 3.
9891 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
9893         PR target/86838
9894         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
9895         * config/aarch64/aarch64-simd.md
9896         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
9897         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
9898         (aarch64_frecpx<mode>): ...this new pattern.
9899         * config/aarch64/aarch64-simd-builtins.def: Remove comment
9900         about aarch64_frecp<FRECP:frecp_suffix><mode>.
9902 2018-08-07  Martin Liska  <mliska@suse.cz>
9904         PR middle-end/83023
9905         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
9906         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
9907         * predict.def (PRED_MALLOC_NONNULL): New predictor.
9908         * doc/extend.texi: Document that malloc attribute adds
9909         hit to compiler.
9911 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
9913         PR target/86785
9914         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9915         Define to speculation_safe_value_not_needed.
9917 2018-08-06  Jeff Law  <law@redhat.com>
9919         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
9920         the vr_values instance to cprop_into_stmt.
9921         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
9922         (cprop_operand): Also query EVRP to determine if OP is a constant.
9924 2018-08-06  Nathan Sidwell  <nathan@acm.org>
9926         * diagnostic.c (diagnostic_report_current_module): Reroll
9927         included-at loop.  Translate text.
9929 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
9931         * function-tests.c (selftest::test_expansion_to_rtl): Call
9932         free_after_compilation.
9934 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9936         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
9938 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
9940         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
9941         loops with memory block operations from getting unrolled.
9943 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
9945         PR target/86807
9946         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9947         Define to speculation_safe_value_not_needed.
9949 2018-08-06  Jeff Law  <law@redhat.com>
9951         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
9952         assert.
9954 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
9956         PR target/86662
9957         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
9958         with all enabled __intN types.
9960         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
9962 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9964         * alias.c (record_set): Check for clobber high.
9965         * cfgexpand.c (expand_gimple_stmt): Likewise.
9966         * combine-stack-adj.c (single_set_for_csa): Likewise.
9967         * combine.c (find_single_use_1): Likewise.
9968         (set_nonzero_bits_and_sign_copies): Likewise.
9969         (get_combine_src_dest): Likewise.
9970         (is_parallel_of_n_reg_sets): Likewise.
9971         (try_combine): Likewise.
9972         (record_dead_and_set_regs_1): Likewise.
9973         (reg_dead_at_p_1): Likewise.
9974         (reg_dead_at_p): Likewise.
9975         * dce.c (deletable_insn_p): Likewise.
9976         (mark_nonreg_stores_1): Likewise.
9977         (mark_nonreg_stores_2): Likewise.
9978         * df-scan.c (df_find_hard_reg_defs): Likewise.
9979         (df_uses_record): Likewise.
9980         (df_get_call_refs): Likewise.
9981         * dwarf2out.c (mem_loc_descriptor): Likewise.
9982         * haifa-sched.c (haifa_classify_rtx): Likewise.
9983         * ira-build.c (create_insn_allocnos): Likewise.
9984         * ira-costs.c (scan_one_insn): Likewise.
9985         * ira.c (equiv_init_movable_p): Likewise.
9986         (rtx_moveable_p): Likewise.
9987         (interesting_dest_for_shprep): Likewise.
9988         * jump.c (mark_jump_label_1): Likewise.
9989         * postreload-gcse.c (record_opr_changes): Likewise.
9990         * postreload.c (reload_cse_simplify): Likewise.
9991         (struct reg_use): Add source expr.
9992         (reload_combine): Check for clobber high.
9993         (reload_combine_note_use): Likewise.
9994         (reload_cse_move2add): Likewise.
9995         (move2add_note_store): Likewise.
9996         * print-rtl.c (print_pattern): Likewise.
9997         * recog.c (decode_asm_operands): Likewise.
9998         (store_data_bypass_p): Likewise.
9999         (if_test_bypass_p): Likewise.
10000         * regcprop.c (kill_clobbered_value): Likewise.
10001         (kill_set_value): Likewise.
10002         * reginfo.c (reg_scan_mark_refs): Likewise.
10003         * reload1.c (maybe_fix_stack_asms): Likewise.
10004         (eliminate_regs_1): Likewise.
10005         (elimination_effects): Likewise.
10006         (mark_not_eliminable): Likewise.
10007         (scan_paradoxical_subregs): Likewise.
10008         (forget_old_reloads_1): Likewise.
10009         * reorg.c (find_end_label): Likewise.
10010         (try_merge_delay_insns): Likewise.
10011         (redundant_insn): Likewise.
10012         (own_thread_p): Likewise.
10013         (fill_simple_delay_slots): Likewise.
10014         (fill_slots_from_thread): Likewise.
10015         (dbr_schedule): Likewise.
10016         * resource.c (update_live_status): Likewise.
10017         (mark_referenced_resources): Likewise.
10018         (mark_set_resources): Likewise.
10019         * rtl.c (copy_rtx): Likewise.
10020         * rtlanal.c (reg_referenced_p): Likewise.
10021         (single_set_2): Likewise.
10022         (noop_move_p): Likewise.
10023         (note_stores): Likewise.
10024         * sched-deps.c (sched_analyze_reg): Likewise.
10025         (sched_analyze_insn): Likewise.
10027 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10029         * cse.c (invalidate_reg): New function extracted from...
10030         (invalidate): ...here.
10031         (canonicalize_insn): Check for clobber high.
10032         (invalidate_from_clobbers): invalidate clobber highs.
10033         (invalidate_from_sets_and_clobbers): Likewise.
10034         (count_reg_usage): Check for clobber high.
10035         (insn_live_p): Likewise.
10036         * cselib.c (cselib_expand_value_rtx_1):Likewise.
10037         (cselib_invalidate_regno): Check for clobber in setter.
10038         (cselib_invalidate_rtx): Pass through setter.
10039         (cselib_invalidate_rtx_note_stores):
10040         (cselib_process_insn): Check for clobber high.
10041         * cselib.h (cselib_invalidate_rtx): Add operand.
10043 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10045         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
10046         (mark_not_eliminable): Likewise.
10047         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
10048         * lra-lives.c (process_bb_lives): Check for clobber high.
10049         * lra.c (new_insn_reg): Remember clobber highs.
10050         (collect_non_operand_hard_regs): Check for clobber high.
10051         (lra_set_insn_recog_data): Likewise.
10052         (add_regs_to_insn_regno_info): Likewise.
10053         (lra_update_insn_regno_info): Likewise.
10055 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10057         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
10058         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
10060 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10062         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
10063         (copy_insn_1): Likewise.
10064         (gen_hard_reg_clobber_high): New gen function.
10065         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
10066         * genemit.c (gen_exp): Likewise.
10067         (gen_emit_seq): Pass through info.
10068         (gen_insn): Check for CLOBBER_HIGH.
10069         (gen_expand): Pass through info.
10070         (gen_split): Likewise.
10071         (output_add_clobbers): Likewise.
10072         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
10073         (remove_clobbers): Likewise.
10074         * rtl.h (gen_hard_reg_clobber_high): New declaration.
10076 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10078         * doc/rtl.texi (clobber_high): Add.
10079         (parallel): Add in clobber high
10080         * rtl.c (rtl_check_failed_code3): Add function.
10081         * rtl.def (CLOBBER_HIGH): Add expression.
10082         * rtl.h (RTL_CHECKC3): Add macro.
10083         (rtl_check_failed_code3): Add declaration.
10084         (XC3EXP): Add macro.
10086 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
10088         PR target/86386
10089         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
10090         cfun->machine->max_used_stack_alignment if needed.
10092 2018-08-04  Martin Sebor  <msebor@redhat.com>
10094         PR tree-optimization/86571
10095         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
10096         NaN output to 4.
10098 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
10100         PR target/86799
10101         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10102         Define.
10104 2018-08-03  Jeff Law  <law@redhat.com>
10106         PR target/86795
10107         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10108         Define to speculation_safe_value_not_needed.
10110 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
10112         * doc/gcov.texi (-x): Remove duplicate "to".
10113         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
10114         (-Wif-not-aligned): Remove duplicate "is".
10115         (-flto): Remove duplicate "the".
10116         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
10117         duplicate "v5.00.b".
10118         (MSP430 Options): Remove duplicate "and" from the description
10119         of "-mgprel-sec=regexp".
10120         (x86 Options): Remove duplicate copies of "vmldLog102" and
10121         vmlsLog104 from description of "-mveclibabi=type".
10123 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
10125         * internal-fn.h (first_commutative_argument): Declare.
10126         * internal-fn.c (first_commutative_argument): New function.
10127         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
10128         restrictions for pattern statements.  Use first_commutative_argument
10129         to look for commutative operands in calls to internal functions.
10131 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
10133         * Makefile.in (wide-int-range.o): New.
10134         * tree-vrp.c: Move all the wide_int_* functions to...
10135         * wide-int-range.cc: ...here.
10136         * tree-vrp.h: Move all the wide_int_* prototypes to...
10137         * wide-int-range.h: ...here.
10139 2018-08-03  Tom de Vries  <tdevries@suse.de>
10141         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
10142         UI_NONE.
10143         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
10144         * except.c (output_function_exception_table): Do early exit if
10145         targetm_common.except_unwind_info (&global_options) == UI_NONE.
10147 2018-08-03  Martin Liska  <mliska@suse.cz>
10149         * predict.c (dump_prediction): Change to 2 digits
10150         in fraction part.
10152 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
10154         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
10155         neon_dup_q to...
10156         (falkor_am_1_gtov_gtov): ... a new insn reservation.
10158 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
10160         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
10161         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
10162         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
10163         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
10164         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
10166 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
10168         * diagnostic-show-locus.c (diagnostic_show_locus): Use
10169         pp_take_prefix when saving the existing prefix.
10170         * diagnostic.c (diagnostic_append_note): Likewise.
10171         * langhooks.c (lhd_print_error_function): Likewise.
10172         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
10173         param's type.  Free the existing prefix.
10174         (pp_take_prefix): New function.
10175         (pretty_printer::pretty_printer): Drop the prefix parameter.
10176         Rename the length parameter to match the comment.
10177         (pretty_printer::~pretty_printer): Free the prefix.
10178         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
10179         parameter.
10180         (struct pretty_printer): Drop the "const" from "prefix" field's
10181         type and clarify memory management.
10182         (pp_set_prefix): Drop the "const" from the 2nd param.
10183         (pp_take_prefix): New decl.
10185 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
10187         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
10188         for word_mode_ok here instead of passing as argument.
10189         (expand_block_compare): Change select_block_compare_mode() call.
10190         (expand_strncmp_gpr_sequence): New function.
10191         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
10193 2018-08-02  Jeff Law  <law@redhat.com>
10195         PR target/86790
10196         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10197         Define to speculation_safe_value_not_needed.
10199         PR target/86784
10200         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10201         Define to speculation_safe_value_not_needed.
10203 2018-08-02  Tom de Vries  <tdevries@suse.de>
10205         PR target/86660
10206         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
10207         function.  Return UI_TARGET unconditionally.
10208         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
10209         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
10211 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
10213         * genemit.c (print_overload_test): Fix typo.
10215 2018-08-02  Richard Biener  <rguenther@suse.de>
10217         PR tree-optimization/86816
10218         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
10219         which checks for value availability before querying it.
10220         (gvn_uses_equal): Use it.
10221         (same_succ_hash): Likewise.
10222         (gimple_equal_p): Likewise.
10224 2018-08-02  Nick Clifton  <nickc@redhat.com>
10226         PR target/86813
10227         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10228         Define to speculation_safe_value_not_needed.
10230         PR target/86810
10231         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10232         Define to speculation_safe_value_not_needed.
10234         PR target/86810
10235         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10236         Define to speculation_safe_value_not_needed.
10238         PR target/86803
10239         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10240         Define to speculation_safe_value_not_needed.
10242         PR target/86797
10243         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10244         Define to speculation_safe_value_not_needed.
10246         PR target/86791
10247         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10248         Define to speculation_safe_value_not_needed.
10250         PR target/86789
10251         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10252         Define to speculation_safe_value_not_needed.
10254         PR target/86787
10255         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10256         Define to speculation_safe_value_not_needed.
10258         PR target/86782
10259         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
10260         speculation_safe_value_not_needed.
10262         PR target/86781
10263         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
10264         to speculation_safe_value_not_needed.
10266 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
10268         * doc/md.texi: Expand the documentation of instruction names
10269         to mention port-local uses.  Document '@' in pattern names.
10270         * read-md.h (overloaded_instance, overloaded_name): New structs.
10271         (mapping): Declare.
10272         (md_reader::handle_overloaded_name): New member function.
10273         (md_reader::get_overloads): Likewise.
10274         (md_reader::m_first_overload): New member variable.
10275         (md_reader::m_next_overload_ptr): Likewise.
10276         (md_reader::m_overloads_htab): Likewise.
10277         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
10278         m_next_overload_ptr and m_overloads_htab.
10279         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
10280         (get_mode_token, get_code_token, get_int_token): New functions.
10281         (map_attr_string): Add an optional argument that passes back
10282         the associated iterator.
10283         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
10284         (md_reader::handle_overloaded_name, add_overload_instance): New
10285         functions.
10286         (apply_iterators): Handle '@' names.  Report an error if '@'
10287         is used without iterators.
10288         (initialize_iterators): Initialize the new iterator_group fields.
10289         * genopinit.c (handle_overloaded_code_for)
10290         (handle_overloaded_gen): New functions.
10291         (main): Use them to print declarations of maybe_code_for_* and
10292         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
10293         * genemit.c (print_overload_arguments, print_overload_test)
10294         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
10295         (main): Use it to print definitions of maybe_code_for_* and
10296         maybe_gen_* functions.
10297         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
10298         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
10299         instead of explicit mode checks.
10300         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
10301         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
10302         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
10303         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
10304         (aarch64_expand_compare_and_swap): Likewise
10305         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
10306         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
10307         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
10308         (aarch64_constant_pool_reload_icode): Delete.
10309         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
10310         instead of aarch64_constant_pool_reload_icode.  Use
10311         code_for_aarch64_reload_mov instead of explicit mode checks.
10312         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
10313         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
10314         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
10315         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
10316         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
10317         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
10318         (aarch64_atomic_load_op_code): Delete.
10319         (aarch64_emit_atomic_load_op): Likewise.
10320         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
10321         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
10322         instead of aarch64_emit_atomic_load_op.
10323         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
10324         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
10325         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
10326         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
10327         character before the pattern name.
10328         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
10329         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
10330         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
10331         (aarch64_frecps<mode>): Likewise.
10332         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
10333         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
10334         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
10335         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
10336         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
10338 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
10340         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
10341         Allow HFmode constants if TARGET_FP_F16INST.
10343 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
10345         PR target/86014
10346         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
10347         No longer check last store for clobber of address register.
10349 2018-08-02  Martin Liska  <mliska@suse.cz>
10351         PR gcov-profile/86817
10352         * gcov.c (process_all_functions): New function.
10353         (main): Call it.
10354         (process_file): Move functions processing to
10355         process_all_functions.
10357 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
10359         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
10360         "const" to the "gimple *" and "rtx_insn *" parameters.
10361         * dumpfile.h (dump_user_location_t::dump_user_location_t):
10362         Likewise.
10363         (dump_location_t::dump_location_t): Likewise.
10365 2018-08-01  Martin Sebor  <msebor@redhat.com>
10367         PR tree-optimization/86650
10368         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
10369         rather than a "gcall *".  Directly pass the data of interest
10370         to percent_K_format, rather than building a temporary CALL_EXPR
10371         to hold it.
10372         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
10373         (gimple_fold_builtin_strncat): Adjust.
10374         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
10375         gcall* argument with gimple*.
10376         * gimple-ssa-warn-restrict.c (check_call): Same.
10377         (wrestrict_dom_walker::before_dom_children): Same.
10378         (builtin_access::builtin_access): Same.
10379         (check_bounds_or_overlap): Same
10380         (maybe_diag_overlap): Same.
10381         (maybe_diag_offset_bounds): Same.
10382         * tree-diagnostic.c (default_tree_printer): Move usage of
10383         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
10384         to this callsite.
10385         * tree-pretty-print.c (percent_K_format): Add argument.
10386         * tree-pretty-print.h: Add argument.
10387         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
10388         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
10389         (maybe_diag_stxncpy_trunc): Same.
10390         (handle_builtin_stxncpy): Same.
10391         (handle_builtin_strcat): Same.
10393 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10395         * match.pd: Optimise pointer range checks.
10397 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10399         PR tree-optimization/86758
10400         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
10401         to remove pattern statements.
10403 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10405         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
10406         result of dfs_enumerate_from when constructing stmt_vec_infos,
10407         instead of additionally calling get_loop_body.
10409 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10411         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
10412         parameter.
10413         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
10414         When creating an iv, assert that the step is not known to be zero.
10415         (vect_setup_realignment): Update call accordingly.
10416         * tree-vect-stmts.c (vectorizable_store): Likewise.
10417         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
10419 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10421         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
10422         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
10423         (vectorizable_reduction): Likewise.
10424         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10425         (vect_detect_hybrid_slp_stmts): Likewise.
10426         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
10428 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
10430         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
10431         (wide_int_set_zero_nonzero_bits): ...this.
10432         (zero_nonzero_bits_from_vr): Rename to...
10433         (vrp_set_zero_nonzero_bits): ...this.
10434         (extract_range_from_multiplicative_op_1): Abstract wide int
10435         code...
10436         (wide_int_range_multiplicative_op): ...here.
10437         (extract_range_from_binary_expr_1): Extract wide int binary
10438         operations into their own functions.
10439         (wide_int_range_lshift): New.
10440         (wide_int_range_can_optimize_bit_op): New.
10441         (wide_int_range_shift_undefined_p): New.
10442         (wide_int_range_bit_xor): New.
10443         (wide_int_range_bit_ior): New.
10444         (wide_int_range_bit_and): New.
10445         (wide_int_range_trunc_mod): New.
10446         (extract_range_into_wide_ints): New.
10447         (vrp_shift_undefined_p): New.
10448         (extract_range_from_multiplicative_op): New.
10449         (vrp_can_optimize_bit_op): New.
10450         * tree-vrp.h (value_range::dump): New.
10451         (wide_int_range_multiplicative_op): New.
10452         (wide_int_range_lshift):New.
10453         (wide_int_range_shift_undefined_p): New.
10454         (wide_int_range_bit_xor): New.
10455         (wide_int_range_bit_ior): New.
10456         (wide_int_range_bit_and): New.
10457         (wide_int_range_trunc_mod): New.
10458         (zero_nonzero_bits_from_bounds): Rename to...
10459         (wide_int_set_zero_nonzero_bits): ...this.
10460         (zero_nonzero_bits_from_vr): Rename to...
10461         (vrp_set_zero_nonzero_bits): ...this.
10462         (range_easy_mask_min_max): Rename to...
10463         (wide_int_range_can_optimize_bit_op): this.
10464         * vr-values.c (simplify_bit_ops_using_ranges): Rename
10465         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
10467 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10469         * tree-vectorizer.h (vect_orig_stmt): New function.
10470         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
10471         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
10472         (vect_create_epilog_for_reduction): Likewise.
10473         (vectorizable_live_operation): Likewise.
10474         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
10475         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
10476         * tree-vect-stmts.c (vectorizable_call): Likewise.
10477         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
10479 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10481         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
10482         argument.
10483         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
10484         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
10485         (vect_transform_loop): Likewise.
10486         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
10488 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10490         * tree-vectorizer.h (vect_schedule_slp): Return void.
10491         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
10492         (vect_schedule_slp): Likewise.
10494 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10496         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
10497         argument.
10498         (vect_transform_loop): Update calls accordingly.  Schedule SLP
10499         instances before the main loop, if any exist.
10501 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10503         PR tree-optimization/86749
10504         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
10505         If the lhs is used in a COND_EXPR, check that it is being used
10506         as the "then" or "else" value.
10508 2018-08-01  Tom de Vries  <tdevries@suse.de>
10510         PR target/86800
10511         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
10512         speculation_safe_value_not_needed.
10514 2018-08-01  Richard Biener  <rguenther@suse.de>
10516         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
10517         as base and offset.
10519 2018-08-01  Martin Liska  <mliska@suse.cz>
10521         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
10522         format how successful transformation is dumped.
10523         (gimple_mod_pow2_value_transform): Likewise.
10524         (gimple_mod_subtract_transform): Likewise.
10525         (gimple_stringops_transform): Likewise.
10527 2018-08-01  Martin Liska  <mliska@suse.cz>
10529         PR value-prof/35543
10530         * value-prof.c (interesting_stringop_to_profile_p):
10531         Simplify the code and add BUILT_IN_MEMMOVE.
10532         (gimple_stringops_transform): Likewise.
10534 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
10536         * config/aarch64/aarch64-simd.md
10537         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
10538         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
10539         use GPI iterator instead of SI mode.
10541 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
10543         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
10544         rs6000_speculation_barrier.
10545         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
10546         new barrier pattern name.
10548 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
10550         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
10551         (speculation_barrier): New insn.
10553 2018-08-01  Richard Biener  <rguenther@suse.de>
10555         PR bootstrap/86724
10556         * graphite.h: Include isl/id.h and isl/space.h to allow build
10557         with ISL 0.20.
10559 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
10561         PR target/86651
10562         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
10563         mode for COFF targets.
10564         * defaults.h (TARGET_COFF): Define.
10565         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
10566         TARGET_COFF): Define.
10567         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
10568         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
10569         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
10571 2018-07-31  Alexandre Oliva <oliva@adacore.com>
10573         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
10574         * gimple-streamer-out.c (output_bb): Save it.
10575         * lto-streamer-in.c (input_struct_function_base): Restore
10576         instance discriminator if available.  Create map on demand.
10577         * lto-streamer-out.c (output_struct_function_base): Save it if
10578         available.
10579         * final.c (decl_to_instance_map): Document LTO strategy.
10581 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
10582             Olivier Hainque  <hainque@adacore.com>
10584         * debug.h (decl_to_instance_map_t): New type.
10585         (decl_to_instance_map): Declare.
10586         (maybe_create_decl_to_instance_map): New inline function.
10587         * final.c (bb_discriminator, last_bb_discriminator): New statics,
10588         to track basic block discriminators.
10589         (final_start_function_1): Initialize them.
10590         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
10591         bb_discriminator.
10592         (decl_to_instance_map): New variable.
10593         (map_decl_to_instance, maybe_set_discriminator): New functions.
10594         (notice_source_line): Set discriminator.
10596 2018-07-31  Ian Lance Taylor  <iant@golang.org>
10598         * targhooks.c (default_have_speculation_safe_value): Add
10599         ATTRIBUTE_UNUSED.
10601 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
10603         * dump-context.h: Include "pretty-print.h".
10604         (dump_context::refresh_dumps_are_enabled): New decl.
10605         (dump_context::emit_item): New decl.
10606         (class dump_context): Add fields "m_test_pp" and
10607         "m_test_pp_flags".
10608         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
10609         (temp_dump_context::get_dumped_text): New decl.
10610         (class temp_dump_context): Add field "m_pp".
10611         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
10612         (dump_context::refresh_dumps_are_enabled): ...and add a test for
10613         m_test_pp.
10614         (set_dump_file): Update for above change.
10615         (set_alt_dump_file): Likewise.
10616         (dump_loc): New overload, taking a pretty_printer *.
10617         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
10618         to any test pretty-printer.
10619         (make_item_for_dump_gimple_stmt): New function, adapted from
10620         optinfo::add_gimple_stmt.
10621         (dump_context::dump_gimple_stmt): Call it, and use the result,
10622         eliminating the direct usage of dump_file and alt_dump_file in
10623         favor of indirectly using them via emit_item.
10624         (make_item_for_dump_gimple_expr): New function, adapted from
10625         optinfo::add_gimple_expr.
10626         (dump_context::dump_gimple_expr): Call it, and use the result,
10627         eliminating the direct usage of dump_file and alt_dump_file in
10628         favor of indirectly using them via emit_item.
10629         (make_item_for_dump_generic_expr): New function, adapted from
10630         optinfo::add_tree.
10631         (dump_context::dump_generic_expr): Call it, and use the result,
10632         eliminating the direct usage of dump_file and alt_dump_file in
10633         favor of indirectly using them via emit_item.
10634         (make_item_for_dump_printf_va): New function, adapted from
10635         optinfo::add_printf_va.
10636         (make_item_for_dump_printf): New function.
10637         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
10638         and use the result, eliminating the direct usage of dump_file and
10639         alt_dump_file in favor of indirectly using them via emit_item.
10640         (make_item_for_dump_dec): New function.
10641         (dump_context::dump_dec): Call it, and use the result,
10642         eliminating the direct usage of dump_file and alt_dump_file in
10643         favor of indirectly using them via emit_item.
10644         (make_item_for_dump_symtab_node): New function, adapted from
10645         optinfo::add_symtab_node.
10646         (dump_context::dump_symtab_node): Call it, and use the result,
10647         eliminating the direct usage of dump_file and alt_dump_file in
10648         favor of indirectly using them via emit_item.
10649         (dump_context::begin_scope): Reimplement, avoiding direct usage
10650         of dump_file and alt_dump_file in favor of indirectly using them
10651         via emit_item.
10652         (dump_context::emit_item): New member function.
10653         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
10654         Set up test pretty-printer on the underlying context.  Call
10655         refresh_dumps_are_enabled.
10656         (temp_dump_context::~temp_dump_context): Call
10657         refresh_dumps_are_enabled.
10658         (temp_dump_context::get_dumped_text): New member function.
10659         (selftest::verify_dumped_text): New function.
10660         (ASSERT_DUMPED_TEXT_EQ): New macro.
10661         (selftest::test_capture_of_dump_calls): Run all tests twice, with
10662         and then without optinfo enabled.  Add uses of
10663         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
10664         * dumpfile.h: Update comment for the dump_* API.
10665         * optinfo-emit-json.cc
10666         (selftest::test_building_json_from_dump_calls): Update for new
10667         param for temp_dump_context ctor.
10668         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
10669         and "m_owned" field.
10670         (optinfo_item::~optinfo_item): Likewise.
10671         (optinfo::add_item): New member function.
10672         (optinfo::emit): Update comment.
10673         (optinfo::add_string): Delete.
10674         (optinfo::add_printf): Delete.
10675         (optinfo::add_printf_va): Delete.
10676         (optinfo::add_gimple_stmt): Delete.
10677         (optinfo::add_gimple_expr): Delete.
10678         (optinfo::add_tree): Delete.
10679         (optinfo::add_symtab_node): Delete.
10680         (optinfo::add_dec): Delete.
10681         * optinfo.h (class dump_context): New forward decl.
10682         (optinfo::add_item): New decl.
10683         (optinfo::add_string): Delete.
10684         (optinfo::add_printf): Delete.
10685         (optinfo::add_printf_va): Delete.
10686         (optinfo::add_gimple_stmt): Delete.
10687         (optinfo::add_gimple_expr): Delete.
10688         (optinfo::add_tree): Delete.
10689         (optinfo::add_symtab_node): Delete.
10690         (optinfo::add_dec): Delete.
10691         (optinfo::add_poly_int): Delete.
10692         (optinfo_item::optinfo_item): Remove "owned" param.
10693         (class optinfo_item): Remove field "m_owned".
10695 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
10697         PR middle-end/86705
10698         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
10699         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
10700         requested variable alignment.
10701         (expand_one_ssa_partition): Likewise.
10702         (expand_one_var): Likewise.
10704 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10706         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
10707         to speculation_safe_value_not_needed.
10709 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10711         * targhooks.h (speculation_safe_value_not_needed): New prototype.
10712         * targhooks.c (speculation_safe_value_not_needed): New function.
10713         * target.def (have_speculation_safe_value): Update documentation.
10714         * doc/tm.texi: Regenerated.
10716 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10718         * config/aarch64/iterators.md (ALLI_TI): New iterator.
10719         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
10720         expand.
10721         (despeculate_copy<ALLI:mode>_insn): New insn.
10722         (despeculate_copyti_insn): New insn.
10723         (despeculate_simple<ALLI:mode>): New insn
10724         (despeculate_simpleti): New insn.
10725         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
10726         function.
10727         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
10728         aarch64_speculation_safe_value.
10729         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
10731 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10733         * config/aarch64/aarch64-speculation.cc: New file.
10734         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
10735         before pass_reorder_blocks.
10736         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
10737         prototype.
10738         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
10739         X14 and X15 when tracking speculation.
10740         * config/aarch64/aarch64.md (register name constants): Add
10741         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
10742         (unspec): Add UNSPEC_SPECULATION_TRACKER.
10743         (speculation_barrier): New insn attribute.
10744         (cmp<mode>): Allow SP in comparisons.
10745         (speculation_tracker): New insn.
10746         (speculation_barrier): Add speculation_barrier attribute.
10747         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
10748         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
10749         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
10751 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10753         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
10754         aarch64_track_speculation is true.
10755         (tb<optab><mode>1): Likewise.
10756         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
10757         generate CB[N]Z when tracking speculation.
10758         (aarch64_split_compare_and_swap): Likewise.
10759         (aarch64_split_atomic_op): Likewise.
10761 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10763         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
10765 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10767         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
10768         (speculation_barrier): New insn.
10770 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10772         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
10773         * config/arm/arm.md (speculation_barrier): New expand.
10774         (speculation_barrier_insn): New pattern.
10776 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10778         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
10779         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
10780         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
10781         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
10782         list.
10783         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
10784         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
10785         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
10786         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
10787         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
10788         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
10789         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
10790         * builtins.c (expand_speculation_safe_value): New function.
10791         (expand_builtin): Call it.
10792         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
10793         * doc/extend.texi: Document __builtin_speculation_safe_value.
10794         * doc/md.texi: Document "speculation_barrier" pattern.
10795         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
10796         TARGET_HAVE_SPECULATION_SAFE_VALUE.
10797         * doc/tm.texi: Regenerated.
10798         * target.def (have_speculation_safe_value, speculation_safe_value): New
10799         hooks.
10800         * targhooks.c (default_have_speculation_safe_value): New function.
10801         (default_speculation_safe_value): New function.
10802         * targhooks.h (default_have_speculation_safe_value): Add prototype.
10803         (default_speculation_safe_value): Add prototype.
10805 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
10807         * dump-context.h (dump_context::dump_loc): New decl.
10808         * dumpfile.c (dump_context::dump_loc): New member function.
10809         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
10810         and dump_gimple_stmt.
10811         (dump_context::dump_gimple_expr_loc): Likewise, using
10812         dump_gimple_expr.
10813         (dump_context::dump_generic_expr_loc): Likewise, using
10814         dump_generic_expr.
10815         (dump_context::dump_printf_loc_va): Likewise, using
10816         dump_printf_va.
10817         (dump_context::begin_scope): Explicitly using the global function
10818         "dump_loc", rather than the member function.
10820 2018-07-31  Martin Sebor  <msebor@redhat.com>
10822         PR tree-optimization/86741
10823         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
10825 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
10827         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
10828         depend on whether prefetch instructions will be emitted or not.
10829         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
10830         will be emitted or not.
10831         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
10832         (TARGET_SETMEM_PFD): New macros.
10834 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10836         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
10837         (NULL_STMT_VEC_INFO): Delete.
10838         (stmt_vec_info::operator*): Likewise.
10839         (stmt_vec_info::operator gimple *): Likewise.
10840         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
10841         of NULL_STMT_VEC_INFO.
10842         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
10843         (vect_reassociating_reduction_p): Likewise.
10844         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
10845         (vectorizable_store): Likewise.
10846         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
10847         (vec_info::free_stmt_vec_infos): Likewise.
10849 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10851         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
10852         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
10854 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10856         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
10857         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
10858         (vec_info::free_stmt_vec_info): New private member functions.
10859         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
10860         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
10861         * tree-parloops.c (gather_scalar_reductions): Remove calls to
10862         set_stmt_vec_info_vec and free_stmt_vec_infos.
10863         * tree-vect-loop.c (_loop_vec_info): Remove call to
10864         set_stmt_vec_info_vec.
10865         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
10866         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
10867         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
10868         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
10869         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
10870         assignments in {vec_info::,}new_stmt_vec_info that are redundant
10871         with the clearing in the xcalloc.
10872         (stmt_vec_info_vec): Delete.
10873         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
10874         (vectorize_loops): Likewise.
10875         (vec_info::~vec_info): Remove argument from call to
10876         free_stmt_vec_infos.
10877         (vec_info::add_stmt): Remove vinfo argument from call to
10878         new_stmt_vec_info.
10880 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10882         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
10883         rather than a gimple stmt.
10884         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
10885         information for pattern statements when passed the original
10886         statement; instead wait to be passed the pattern statement itself.
10887         Don't call set_vinfo_for_stmt here.
10888         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
10889         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
10890         stmt_vec_infos here.
10891         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
10892         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
10893         stmt_vec_infos entry.
10895 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10897         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
10898         * tree-vectorizer.c (vec_info::replace_stmt): New function.
10899         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
10900         * tree-vect-stmts.c (vectorizable_call): Likewise.
10901         (vectorizable_simd_clone_call): Likewise.
10903 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10905         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
10906         * tree-vectorizer.c (vec_info::remove_stmt): New function.
10907         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
10908         * tree-vect-loop.c (vect_transform_loop): Likewise.
10909         * tree-vect-slp.c (vect_schedule_slp): Likewise.
10910         * tree-vect-stmts.c (vect_remove_stores): Likewise.
10912 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10914         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
10915         (vect_dr_stmt): Delete.
10916         * tree-vectorizer.c (vec_info::lookup_dr): New function.
10917         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
10918         of DR_VECT_AUX.
10919         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10920         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
10921         (vect_verify_datarefs_alignment, vect_peeling_supportable)
10922         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
10923         (vect_analyze_data_refs): Likewise.
10924         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
10925         argument.
10926         (vect_find_same_alignment_drs): Likewise.
10927         (vect_slp_analyze_node_dependences): Update calls accordingly.
10928         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
10929         instead of DR_VECT_AUX.
10930         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
10931         of a vector data references.  Use vec_info::lookup_dr instead of
10932         DR_VECT_AUX.
10933         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
10934         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
10935         instead of DR_VECT_AUX.
10937 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10939         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
10940         dr_vec_info.
10941         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
10942         accordingly.
10943         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10944         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10945         (vect_gen_prolog_loop_niters): Likewise.
10947 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10949         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
10950         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
10951         (vect_known_alignment_in_bytes, vect_dr_behavior)
10952         (vect_get_scalar_dr_size): Take references as dr_vec_infos
10953         instead of data_references.  Update calls to other routines for
10954         which the same change has been made.
10955         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
10956         dr_vec_infos instead of stmt_vec_infos.
10957         (vect_analyze_data_ref_dependence): Update call accordingly.
10958         (vect_slp_analyze_data_ref_dependence)
10959         (vect_record_base_alignments): Use DR_VECT_AUX.
10960         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
10961         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
10962         (vector_alignment_reachable_p, vect_get_data_access_cost)
10963         (vect_peeling_supportable, vect_analyze_group_access_1)
10964         (vect_analyze_group_access, vect_analyze_data_ref_access)
10965         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
10966         (vect_compile_time_alias, vect_small_gap_p)
10967         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
10968         (vect_supportable_dr_alignment): Take references as dr_vec_infos
10969         instead of data_references.  Update calls to other routines for
10970         which the same change has been made.
10971         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
10972         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
10973         (vect_slp_analyze_and_verify_node_alignment)
10974         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
10975         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
10976         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
10977         above changes.
10978         (_vect_peel_info::dr): Replace with...
10979         (_vect_peel_info::dr_info): ...this new field.
10980         (vect_peeling_hash_get_most_frequent)
10981         (vect_peeling_hash_choose_best_peeling): Update accordingly.
10982         (vect_peeling_hash_get_lowest_cost):
10983         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
10984         routines for which the same change has been made.
10985         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
10986         data_reference.
10987         * tree-vect-loop-manip.c (get_misalign_in_elems)
10988         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
10989         above changes.
10990         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10991         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
10992         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
10993         (get_group_load_store_type, get_negative_load_store_type)
10994         (vect_get_data_ptr_increment, vectorizable_store)
10995         (vectorizable_load): Likewise.
10996         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
10997         Update calls to other routines for which the same change has been made.
10999 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11001         * tree-vectorizer.h (vec_info::move_dr): New member function.
11002         (dataref_aux): Rename to...
11003         (dr_vec_info): ...this and add "dr" and "stmt" fields.
11004         (_stmt_vec_info::dr_aux): Update accordingly.
11005         (_stmt_vec_info::data_ref_info): Delete.
11006         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
11007         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
11008         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
11009         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
11010         of data_ref.
11011         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
11012         (STMT_VINFO_DR_INFO): New macro.
11013         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
11014         (set_dr_misalignment): Update after rename of dataref_aux.
11015         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
11016         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
11017         initialization of STMT_VINFO_DATA_REF.
11018         * tree-vectorizer.c (vec_info::move_dr): New function.
11019         * tree-vect-patterns.c (vect_recog_bool_pattern)
11020         (vect_recog_mask_conversion_pattern)
11021         (vect_recog_gather_scatter_pattern): Use it.
11022         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
11023         the "dr" and "stmt" fields of dr_vec_info instead of
11024         STMT_VINFO_DATA_REF.
11026 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11028         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
11029         (is_pattern_stmt_p): Use it.
11030         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
11031         on pattern statements.
11033 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11035         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
11036         original stmt as a stmt_vec_info rather than a gimple stmt.
11037         (vect_pattern_recog_1): Take the statement directly as a
11038         stmt_vec_info, rather than via a gimple_stmt_iterator.
11039         Update call to vect_mark_pattern_stmts.
11040         (vect_pattern_recog): Update calls accordingly.
11042 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11044         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
11045         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
11046         a vect_def_type for the first argument.
11047         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
11048         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
11049         operand if it isn't defined by a vectorized statement.
11050         (vect_build_gather_load_calls): Remove the mask_dt argument and
11051         update calls to vect_get_vec_def_for_stmt_copy.
11052         (vectorizable_bswap): Likewise the dt argument.
11053         (vectorizable_call): Update calls to vectorizable_bswap and
11054         vect_get_vec_def_for_stmt_copy.
11055         (vectorizable_simd_clone_call, vectorizable_assignment)
11056         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
11057         (vectorizable_comparison): Update calls to
11058         vect_get_vec_def_for_stmt_copy.
11059         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
11060         vect_is_simple_use.
11061         (vect_get_loop_based_defs): Remove dt argument and update call
11062         to vect_get_vec_def_for_stmt_copy.
11063         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
11064         and vect_get_vec_def_for_stmt_copy.
11065         (vectorizable_load): Update calls to vect_build_gather_load_calls
11066         and vect_get_vec_def_for_stmt_copy.
11067         * tree-vect-loop.c (vect_create_epilog_for_reduction)
11068         (vectorizable_reduction, vectorizable_live_operation): Update calls
11069         to vect_get_vec_def_for_stmt_copy.
11071 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11073         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
11074         and gimple stmt arguments with a stmt_vec_info.
11075         (vect_record_base_alignments): Update calls accordingly.
11076         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
11077         and gimple stmt arguments with a stmt_vec_info.
11078         (vect_build_slp_tree_1): Remove vinfo argument and update call
11079         to vect_record_max_nunits.
11080         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
11081         and vect_record_max_nunits.
11083 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11085         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
11086         file and take a stmt_vec_info instead of a gimple stmt.
11087         (supportable_widening_operation, vect_finish_replace_stmt)
11088         (vect_finish_stmt_generation, vect_get_store_rhs)
11089         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
11090         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
11091         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
11092         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
11093         (vect_create_data_ref_ptr, bump_vector_ptr)
11094         (vect_permute_store_chain, vect_setup_realignment)
11095         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
11096         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
11097         (vectorizable_reduction, vectorizable_induction)
11098         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
11099         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
11100         than gimple stmts as arguments.
11101         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
11102         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
11103         (can_group_stmts_p, vect_check_gather_scatter)
11104         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
11105         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
11106         (vect_permute_load_chain, vect_shift_permute_load_chain)
11107         (vect_transform_grouped_load)
11108         (vect_record_grouped_load_vectors): Likewise.
11109         * tree-vect-loop.c (vect_fixup_reduc_chain)
11110         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
11111         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
11112         (vectorizable_reduction, vectorizable_induction)
11113         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
11114         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
11115         (vect_get_load_store_mask): Likewise.
11116         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
11117         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
11118         * tree-vect-stmts.c (vect_mark_relevant)
11119         (is_simple_and_all_uses_invariant)
11120         (exist_non_indexing_operands_for_use_p, process_use)
11121         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
11122         (vect_get_vec_def_for_operand, vect_get_vec_defs)
11123         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
11124         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
11125         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
11126         (get_negative_load_store_type, get_load_store_type)
11127         (vect_check_load_store_mask, vect_check_store_rhs)
11128         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
11129         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11130         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11131         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11132         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
11133         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
11134         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
11135         (supportable_widening_operation): Likewise.
11137 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11139         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
11140         a stmt_vec_info instead of a gcall.
11141         (vect_check_gather_scatter): Update call accordingly.
11142         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
11143         of a gphi.
11144         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
11145         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
11146         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
11147         instead of a gimple stmt.
11148         (vect_transform_loop): Update calls accordingly.
11149         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
11150         stmt_vec_infos instead of gimple stmts.
11151         (vect_analyze_slp_instance): Update use accordingly.
11152         * tree-vect-stmts.c (read_vector_array, write_vector_array)
11153         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
11154         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
11155         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
11156         (vect_gen_widened_results_half, vect_get_loop_based_defs)
11157         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
11158         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
11159         down to subroutines.
11161 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11163         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
11164         of the worklist from a vector of gimple stmts to a vector of
11165         stmt_vec_infos.
11166         * tree-vect-stmts.c (vect_mark_relevant, process_use)
11167         (vect_mark_stmts_to_be_vectorized): Likewise
11169 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11171         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
11172         statement before passing it to vect_analyze_stmt.
11173         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
11174         the chain of phi vector definitions.  Track the exit phi via its
11175         stmt_vec_info.
11176         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
11177         STMT_VINFO_REDUC_DEF.
11178         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
11179         stmt_vec_infos to handle the statement chains.
11180         (vect_get_slp_defs): Record the first statement in the node
11181         using a stmt_vec_info.
11182         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
11183         statements here and pass their stmt_vec_info down to subroutines.
11184         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
11185         down to vect_finish_stmt_generation.
11186         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
11187         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
11188         stmt_vec_infos to subroutines.
11189         (vect_remove_stores): Use stmt_vec_infos to handle the statement
11190         chains.
11192 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11194         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
11195         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
11196         (vect_permute_store_chain, vect_setup_realignment)
11197         (vect_permute_load_chain, vect_shift_permute_load_chain)
11198         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
11199         stmts internally, and when passing values to other vectorizer routines.
11200         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
11201         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
11202         (vect_analyze_loop_operations, get_initial_def_for_reduction)
11203         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
11204         (vectorizable_reduction, vectorizable_induction)
11205         (vectorizable_live_operation, vect_transform_loop_stmt)
11206         (vect_transform_loop): Likewise.
11207         * tree-vect-patterns.c (vect_reassociating_reduction_p)
11208         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
11209         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
11210         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11211         (vect_slp_analyze_node_operations_1): Likewise.
11212         * tree-vect-stmts.c (vect_mark_relevant, process_use)
11213         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
11214         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
11215         (vect_finish_stmt_generation_1, get_group_load_store_type)
11216         (get_load_store_type, vect_build_gather_load_calls)
11217         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11218         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11219         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11220         (vectorizable_store, vectorizable_load, vectorizable_condition)
11221         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
11222         (supportable_widening_operation): Likewise.
11223         (vect_get_vector_types_for_stmt): Likewise.
11224         * tree-vectorizer.h (vect_dr_behavior): Likewise.
11226 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11228         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11229         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
11230         (vect_permute_store_chain, vect_permute_load_chain)
11231         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
11232         repeated stmt_vec_info lookups.
11233         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
11234         (vect_update_ivs_after_vectorizer): Likewise.
11235         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
11236         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
11237         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
11238         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11239         (vect_bb_slp_scalar_cost): Likewise.
11240         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
11242 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11244         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
11245         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
11246         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
11247         to get gassigns and gcalls, rather than operating on generc gimple
11248         stmts.
11249         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
11250         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
11251         (vectorizable_load, vect_analyze_stmt): Likewise.
11252         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
11254 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11256         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
11257         return stmt_vec_infos rather than gimple stmts.  Do not accept
11258         null arguments.
11259         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
11260         of a gimple stmt.
11261         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
11262         Update use of get_later_stmt.
11263         (vect_get_constant_vectors): Update call accordingly.
11264         (vect_schedule_slp_instance): Likewise
11265         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
11266         (vect_slp_analyze_instance_dependence): Likewise.
11267         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
11269 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11271         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
11272         (stmt_info_for_cost::stmt_info): ...this new field.
11273         (add_stmt_costs): Update accordingly.
11274         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
11275         (vect_get_known_peeling_cost): Likewise.
11276         (vect_estimate_min_profitable_iters): Likewise.
11277         * tree-vect-stmts.c (record_stmt_cost): Likewise.
11279 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11281         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
11282         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11283         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
11284         accordingly.
11285         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
11287 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11289         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
11290         a gimple stmt to a stmt_vec_info.
11291         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
11293 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11295         * tree-vectorizer.h (vec_info::grouped_stores): Change from
11296         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11297         (_loop_vec_info::reduction_chains): Likewise.
11298         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
11299         accordingly.
11300         * tree-vect-slp.c (vect_analyze_slp): Likewise.
11302 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11304         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
11305         a gimple stmt to a stmt_vec_info.
11306         (_stmt_vec_info::next_element): Likewise.
11307         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
11308         (vect_slp_analyze_and_verify_node_alignment)
11309         (vect_analyze_group_access_1, vect_analyze_group_access)
11310         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
11311         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
11312         (vect_supportable_dr_alignment): Update accordingly.
11313         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
11314         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
11315         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
11316         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
11317         * tree-vect-slp.c (vect_build_slp_tree_1)
11318         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
11319         (vect_split_slp_store_group, vect_analyze_slp_instance)
11320         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
11321         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
11322         (get_group_load_store_type, get_load_store_type)
11323         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
11324         (vect_transform_stmt, vect_remove_stores): Likewise.
11326 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11328         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
11329         than a gimple stmt.
11330         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11331         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
11332         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
11333         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
11334         (vector_alignment_reachable_p, vect_get_data_access_cost)
11335         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
11336         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
11337         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
11338         (vect_analyze_group_access_1, vect_analyze_group_access)
11339         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
11340         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
11341         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
11342         result of vect_dr_stmt and use the stmt_vec_info instead of
11343         the associated gimple stmt.
11344         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
11345         (vect_gen_prolog_loop_niters): Likewise.
11346         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
11348 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11350         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
11351         to a vec<stmt_vec_info>.
11352         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
11353         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
11354         vec<stmt_vec_info>.
11355         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
11356         to a vec<stmt_vec_info>.
11357         (bst_traits::value_type, bst_traits::value_type): Likewise.
11358         (bst_traits::hash): Update accordingly.
11359         (vect_get_and_check_slp_defs): Change the stmts parameter from
11360         a vec<gimple *> to a vec<stmt_vec_info>.
11361         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
11362         (vect_build_slp_tree): Likewise.
11363         (vect_build_slp_tree_2): Likewise.  Update uses of
11364         SLP_TREE_SCALAR_STMTS.
11365         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
11366         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
11367         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
11368         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
11369         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
11370         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
11371         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
11372         (vect_get_constant_vectors, vect_get_slp_defs)
11373         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
11374         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
11375         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
11376         instead of gimple stmts.
11377         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
11378         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
11379         (vect_slp_analyze_instance_dependence): Update uses of
11380         SLP_TREE_SCALAR_STMTS.
11381         (vect_slp_analyze_and_verify_node_alignment): Likewise.
11382         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
11383         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
11384         (get_initial_defs_for_reduction): Likewise.
11385         (vect_create_epilog_for_reduction): Likewise.
11386         (vectorize_fold_left_reduction): Likewise.
11387         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
11388         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
11389         (can_vectorize_live_stmts): Likewise.
11391 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11393         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
11394         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11395         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
11396         than gimple stmts.
11397         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
11398         of a gimple stmt.
11399         (gather_scalar_reductions): Update after above interface changes.
11400         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
11401         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
11402         than gimple stmts.
11403         (vect_force_simple_reduction): Likewise.
11404         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
11405         LOOP_VINFO_REDUCTIONS.
11406         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11408 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11410         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
11411         a gimple stmt to a stmt_vec_info.
11412         * tree-vect-loop.c (vect_active_double_reduction_p)
11413         (vect_force_simple_reduction, vectorizable_reduction): Update
11414         accordingly.
11416 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11418         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
11419         vec<gimple *> to a vec<stmt_vec_info>.
11420         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
11421         the reduction_phis argument from a vec<gimple *> to a
11422         vec<stmt_vec_info>.
11423         (vectorizable_reduction): Likewise the phis local variable that
11424         is passed to vect_create_epilog_for_reduction.  Update for new type
11425         of SLP_TREE_VEC_STMTS.
11426         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
11427         (vectorizable_live_operation): Likewise.
11428         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
11429         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
11431 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11433         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
11434         a gimple stmt to a stmt_vec_info.
11435         (vectorizable_condition, vectorizable_live_operation)
11436         (vectorizable_reduction, vectorizable_induction): Pass back the
11437         vectorized statement as a stmt_vec_info.
11438         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
11439         use of STMT_VINFO_VEC_STMT.
11440         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
11441         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
11442         as stmt_vec_infos rather than gimple stmts.
11443         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
11444         to a stmt_vec_info.
11445         (vectorizable_live_operation): Likewise.
11446         (vectorizable_reduction, vectorizable_induction): Likewise,
11447         updating use of STMT_VINFO_VEC_STMT.
11448         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
11449         of STMT_VINFO_VEC_STMT.
11450         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
11451         (vectorizable_simd_clone_call, vectorizable_conversion)
11452         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11453         (vectorizable_store, vectorizable_load, vectorizable_condition)
11454         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
11455         from a gimple stmt to a stmt_vec_info.
11456         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
11457         pointer to a stmt_vec_info to the vectorizable_* routines.
11459 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11461         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
11462         a gimple stmt to a stmt_vec_info.
11463         (is_pattern_stmt_p): Update accordingly.
11464         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
11465         (vect_record_grouped_load_vectors): Likewise.
11466         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
11467         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
11468         (vect_model_reduction_cost): Likewise.
11469         (vect_create_epilog_for_reduction): Likewise.
11470         (vectorizable_reduction, vectorizable_induction): Likewise.
11471         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11472         Return the stmt_vec_info for the pattern statement.
11473         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
11474         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
11475         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
11476         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
11477         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
11478         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
11479         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
11480         (free_stmt_vec_info, vect_is_simple_use): Likewise.
11482 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11484         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
11485         (vect_finish_stmt_generation): Likewise.
11486         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
11487         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
11488         (vect_build_gather_load_calls): Use the return value of the above
11489         functions instead of a separate call to vinfo_for_stmt.  Use narrow
11490         scopes for the input gimple stmt and wider scopes for the associated
11491         stmt_vec_info.  Use vec_info::lookup_def when setting these
11492         stmt_vec_infos from an SSA_NAME definition.
11493         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11494         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11495         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11496         (vectorizable_store, vectorizable_load, vectorizable_condition)
11497         (vectorizable_comparison): Likewise.
11498         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
11499         (vectorizable_reduction): Likewise.
11501 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11503         * tree-vectorizer.h (vect_is_simple_use): Add an optional
11504         stmt_vec_info * parameter before the optional gimple **.
11505         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
11506         (process_use, vect_get_vec_def_for_operand_1): Update callers.
11507         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
11508         * tree-vect-loop.c (vectorizable_reduction): Likewise.
11509         (vectorizable_live_operation): Likewise.
11510         * tree-vect-patterns.c (type_conversion_p): Likewise.
11511         (vect_look_through_possible_promotion): Likewise.
11512         (vect_recog_rotate_pattern): Likewise.
11513         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
11515 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11517         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
11518         a typedef to a wrapper class.
11519         (NULL_STMT_VEC_INFO): New macro.
11520         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
11521         (stmt_vec_info::operator*): New function.
11522         (stmt_vec_info::operator gimple *): Likewise.
11523         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
11524         (add_stmt_costs): Likewise.
11525         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
11526         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
11527         (vect_get_known_peeling_cost): Likewise.
11528         (vect_estimate_min_profitable_iters): Likewise.
11529         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11530         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
11531         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
11532         (vectorizable_store, free_stmt_vec_infos): Likewise.
11533         (new_stmt_vec_info): Change return type of xcalloc to
11534         _stmt_vec_info *.
11536 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11538         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
11539         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
11540         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
11541         a single_imm_use-based sequence.
11542         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
11544 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11546         * tree-vectorizer.h (vec_info::lookup_def): Declare.
11547         * tree-vectorizer.c (vec_info::lookup_def): New function.
11548         * tree-vect-patterns.c (vect_get_internal_def): Use it.
11549         (vect_widened_op_tree): Likewise.
11550         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
11551         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
11552         (vectorizable_reduction): Likewise.
11553         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
11554         of a gimple *.
11555         (vect_is_slp_reduction): Update calls accordingly.  Use
11556         vec_info::lookup_def.
11557         (vect_is_simple_reduction): Likewise
11558         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
11560 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11562         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
11563         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
11564         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
11565         of vinfo_for_stmt.
11566         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
11567         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
11568         (vect_update_vf_for_slp, vect_analyze_loop_operations)
11569         (vect_is_slp_reduction, vectorizable_induction)
11570         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
11571         * tree-vect-patterns.c (vect_init_pattern_stmt):
11572         (vect_determine_min_output_precision_1, vect_determine_precisions)
11573         (vect_pattern_recog): Likewise.
11574         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
11575         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
11576         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
11577         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
11578         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
11579         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
11580         info field from a loop to a loop_vec_info.
11582 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11584         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
11585         (vec_info::add_stmt): Declare.
11586         * tree-vectorizer.c (vec_info::add_stmt): New function.
11587         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
11588         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
11589         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
11590         (vectorizable_induction): Likewise.
11591         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
11592         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
11593         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
11594         (vectorizable_load): Likewise.
11595         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11596         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
11597         (vect_recog_gather_scatter_pattern): Likewise.
11598         (append_pattern_def_seq): Likewise.  Remove a check that is
11599         performed by add_stmt itself.
11601 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11603         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
11604         which make_ssa_name was called with new_stmt before new_stmt
11605         had been created.
11607 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11609         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
11610         split out from...
11611         (vect_is_slp_reduction): ...here...
11612         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
11613         that are already known to be false.
11615 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11617         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
11618         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
11619         STMT_VINFO_NUM_SLP_USES when it's true.
11620         (vect_free_slp_instance): Add a final_p parameter and pass it to
11621         vect_free_slp_tree.
11622         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
11623         (vect_analyze_slp_instance): Likewise.
11624         (vect_slp_analyze_operations): Likewise.
11625         (vect_slp_analyze_bb_1): Likewise.
11626         * tree-vectorizer.c (vec_info): Likewise.
11627         * tree-vect-loop.c (vect_transform_loop): Likewise.
11629 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11631         * tree-vect-loop.c (vectorizable_reduction): Assert that the
11632         function is not called for second and subsequent members of
11633         a reduction group.
11635 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11637         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
11638         cases for nested loops from here to ...
11639         (vect_create_epilog_for_reduction): ...here.  Only call
11640         vect_is_simple_use for inner-loop reductions.
11642 2018-07-31  Martin Liska  <mliska@suse.cz>
11644         PR gcov-profile/85338
11645         PR gcov-profile/85350
11646         PR gcov-profile/85372
11647         * profile.c (struct location_triplet): New.
11648         (struct location_triplet_hash): Likewise.
11649         (output_location): Do not output a BB that
11650         is already recorded for a line.
11651         (branch_prob): Use streamed_locations.
11653 2018-07-31  Martin Liska  <mliska@suse.cz>
11655         PR gcov-profile/85370
11656         * coverage.c (coverage_begin_function): Do not mark target
11657         clones as artificial functions.
11659 2018-07-31  Martin Liska  <mliska@suse.cz>
11661         PR gcov-profile/83813
11662         PR gcov-profile/84758
11663         PR gcov-profile/85217
11664         PR gcov-profile/85332
11665         * profile.c (branch_prob): Do not record GOTO expressions
11666         for GIMPLE statements which locations are already streamed.
11668 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11670         * gcc.c (handle_spec_function): Accept a soft_matched_part
11671         argument, as do_spec_1.  Pass it down to ...
11672         (eval_spec_function): Accept a soft_matched_part argument,
11673         and pass it down to ...
11674         (do_spec_2): Accept a soft_matched_part argument, and pass
11675         it down to do_spec_1.
11676         (do_spec_1): Pass soft_matched_part to handle_spec_function.
11677         (handle_braces): Update call to handle_spec_function.
11678         (driver::set_up_specs): Update calls to do_spec_2.
11679         (compare_debug_dump_opt_spec_function): Likewise.
11680         (compare_debug_self_opt_spec_function): Likewise.
11682 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11684         * common.opt (nolibc): New option.
11685         * doc/invoke.texi (Link Options): Document it.
11686         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
11687         * config/alpha/linux.h: Likewise.
11688         * config/arc/elf.h: Likewise.
11689         * config/arm/uclinux-elf.h: Likewise.
11690         * config/arm/unknown-elf.h: Likewise.
11691         * config/avr/avrlibc.h: Likewise.
11692         * config/bfin/bfin.h: Likewise.
11693         * config/bfin/linux.h: Likewise.
11694         * config/bfin/uclinux.h: Likewise.
11695         * config/darwin.h: Likewise.
11696         * config/darwin10.h: Likewise.
11697         * config/darwin12.h: Likewise.
11698         * config/gnu-user.h: Likewise.
11699         * config/lm32/uclinux-elf.h: Likewise.
11700         * config/pa/pa-hpux11.h: Likewise.
11701         * config/pa/pa64-hpux.h: Likewise.
11702         * config/sparc/sparc.h: Likewise.
11704 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11706         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
11707         undefined variables.
11709 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
11711         PR target/86640
11712         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
11713         instead of GEN_INT.
11715 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11717         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
11718         terminated string literal.
11720 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
11722         PR rtl-optimization/85160
11723         * combine.c (is_just_move): New function.
11724         (try_combine): Allow combining two instructions into two if neither of
11725         the original instructions was a move.
11727 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
11729         PR target/86673
11730         * doc/extend.texi (Global Register Variables): Discourage use of type
11731         qualifiers.
11732         (Local Register Variables): Likewise.
11734 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
11736         PR tree-optimization/86506
11737         * hwint.h (ceil_log2): Resync with hwint.c implementation.
11739 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
11741         PR target/86547
11742         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
11743         hard_regno, make sure no insn between `from` and `to` clobbers it.
11745 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
11746             Tom de Vries  <tdevries@suse.de>
11748         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
11749         (PTX_DEFAULT_RUNTIME_DIM): ... this.
11750         (nvptx_goacc_validate_dims): Set default worker and gang dims to
11751         PTX_DEFAULT_RUNTIME_DIM.
11752         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
11754 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
11756         * config/pa/pa.c (pa_output_addr_vec): Align address table.
11757         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
11758         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
11760 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
11762         * config/rs6000/constraints.md (wG constraint): Delete, no longer
11763         used.
11764         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
11765         predicate to reflect toc fusion has been deleted.
11766         (toc_fusion_mem_raw): Delete, no longer used.
11767         (toc_fusion_mem_wrapped): Likewise.
11768         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
11769         fusion mask bit.
11770         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
11771         Delete, no longer used.
11772         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
11773         meant to be used for toc fusion.
11774         (rs6000_debug_print_mode): Delete toc fusion debugging.
11775         (rs6000_debug_reg_global): Likewise.
11776         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
11777         fusion and secondary reload support that were never used.
11778         (rs6000_option_override_internal): Delete TOC fusion, that was only
11779         partially defined, and it did not work unless you also used the
11780         -mcmodel= switch.
11781         (rs6000_legitimate_address_p): Delete TOC fusion support.
11782         (rs6000_opt_masks): Likewise.
11783         (fusion_wrap_memory_address): Delete function, no longer used.
11784         (fusion_split_address); Delete TOC fusion support.
11785         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
11786         longer used with toc fusion being deleted.
11787         (TARGET_TOC_FUSION_FP): Likewise.
11788         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
11789         UNSPEC.
11790         (toc fusion spliter): Delete TOC fusion support.
11791         (toc_fusionload_<mode>): Likewise.
11792         (toc_fusionload_di): Likewise.
11793         (fusion_gpr_load_<mode>): Delete generator function, this insn no
11794         longer needs to be named.  Rename predicate to delete TOC fusion.
11795         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
11796         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
11797         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
11798         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
11799         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
11801 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11803         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
11804         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
11805         __int128 in built-in function prototypes.
11806         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
11807         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
11809 2018-07-27  Martin Sebor  <msebor@redhat.com>
11811         PR tree-optimization/86696
11812         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
11813         types, including enums.
11814         (handle_char_store): Be prepared for the above function to fail.
11816 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
11818         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
11819         when optimization level is lower than 2 or optimize for size.
11821 2018-07-26  Martin Sebor  <msebor@redhat.com>
11823         PR tree-optimization/86043
11824         PR tree-optimization/86042
11825         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
11826         (get_string_cst_length): Rename...
11827         (get_min_string_length): ...to this.  Add argument.
11828         (handle_char_store): Extend to handle multi-character stores by
11829         MEM_REF.
11830         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
11831         * tree.h (initializer_zerop): Add argument.
11833 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
11835         PR middle-end/86660
11836         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
11837         declare target to variables if they have always,{to,from,tofrom} map
11838         kinds.
11840 2018-07-26  Martin Liska  <mliska@suse.cz>
11842         PR lto/86548
11843         * lto-wrapper.c: Add linker_output as prefix
11844         for ltrans_output_file.
11846 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
11848         PR rtl-optimization/85805
11849         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
11850         value for hard registers if that was written in the same mode.
11852 2018-07-26  Martin Liska  <mliska@suse.cz>
11854         PR gcov-profile/86536
11855         * gcov.c (format_gcov): Use printf format %.*f directly
11856         and do not handle special values.
11858 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11860         * common/config/arc/arc-common.c (arc_option_optimization_table):
11861         Update default optimizations for size.
11863 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11865         * config/arc/arc.md (movsf_insn): Add short instruction selection.
11866         * config/arc/constraints.md (CfZ): New constraint.
11867         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
11868         (subsf3_fpu): Likewise.
11869         (cmpsf_fpu): Likewise.
11870         (cmpsf_fpu_uneq): Likewise.
11872 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11874         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
11875         accesses as well.
11876         (arc_is_uncached_mem_p): uncached applies to both the variable and
11877         the pointer.
11879 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11881         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
11882         register names.
11884 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
11886         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
11887         field "m_scopes" from vec to auto_vec.
11889 2018-07-25  Martin Liska  <mliska@suse.cz>
11891         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
11892         return type.
11894 2018-07-25  Richard Biener  <rguenther@suse.de>
11896         PR debug/86654
11897         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
11898         special wrt context_die late.
11899         (gen_subprogram_die): Re-use DIEs in local scope.
11901 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
11903         PR tree-optimization/86644
11904         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
11906 2018-07-25  Martin Liska  <mliska@suse.cz>
11908         PR middle-end/86645
11909         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
11910         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
11912 2018-07-25  Martin Liska  <mliska@suse.cz>
11914         PR sanitizer/79635
11915         * params.def: Explain ASan abbreviation and provide
11916         a documentation link.
11918 2018-07-24  Martin Sebor  <msebor@redhat.com>
11920         PR tree-optimization/86622
11921         PR tree-optimization/86532
11922         * builtins.h (string_length): Declare.
11923         * builtins.c (c_strlen): Correct handling of non-constant offsets.
11924         (check_access): Be prepared for non-constant length ranges.
11925         (string_length): Make extern.
11926         * expr.c (string_constant): Only handle the minor non-constant
11927         array index.  Use string_constant to compute the length of
11928         a generic string constant.
11930 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
11932         PR tree-optimization/86618
11933         * tree-vect-stmts.c (vectorizable_call): Don't take the address
11934         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
11936 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
11938         PR tree-optimization/86636
11939         * json.cc (json::object::set): Fix comment.  Add assertions.
11940         (json::array::append): Move here from json.h.  Add comment and an
11941         assertion.
11942         (json::string::string): Likewise.
11943         * json.h (json::array::append): Move to json.cc.
11944         (json::string::string): Likewise.
11945         * optinfo-emit-json.cc
11946         (optrecord_json_writer::impl_location_to_json): Assert that we
11947         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
11948         wrapper around it.  Expand the location once, rather than three
11949         times.
11950         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
11951         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
11952         wrappers.
11953         (optrecord_json_writer::optinfo_to_json): Likewise, in four
11954         places.  Fix some overlong lines.
11956 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
11958         * config/aarch64/aarch64-simd.md
11959         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
11960         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
11961         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
11962         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
11963         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
11964         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
11965         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
11966         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
11967         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
11969 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
11971         PR middle-end/86627
11972         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
11973         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
11974         and abs_d == d, do the power of two handling if profitable.
11976 2018-07-24  Richard Biener  <rguenther@suse.de>
11978         * match.pd: Add BIT_FIELD_REF canonicalizations.
11980 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11982         PR c/86617
11983         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
11985 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11987         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
11988         terminated STRING_CST object.
11990 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11992         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
11994 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
11996         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
11997         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
11998         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
11999         the elements into a register.
12000         (rs6000_split_v4si_init_di_reg): Delete.
12001         (rs6000_split_v4si_init): Delete.
12002         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
12003         (vsx_init_v4si): Rewrite as a define_expand.
12005 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
12007         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
12008         zero_extend argument from memory): New.
12010 2018-07-22  Martin Sebor  <msebor@redhat.com>
12012         PR bootstrap/86621
12013         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
12014         diagnosing calls with unknown arguments unless -Walloca-larger-than
12015         is restricted to less than PTRDIFF_MAX bytes.
12017 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
12019         * doc/gcov.texi (Invoking Gcov): Editorial changes.
12021 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
12023         * pretty-print.c (text_info::set_location): Remove redundant
12024         "line_table" parameter from call to rich_location::set_range.
12026 2018-07-20  Martin Sebor  <msebor@redhat.com>
12028         PR middle-end/82063
12029         * builtins.c (expand_builtin_alloca): Adjust.
12030         * calls.c (alloc_max_size): Simplify.
12031         * cgraphunit.c (cgraph_node::expand): Adjust.
12032         * common.opt (larger_than_size, warn_frame_larger_than): Remove
12033         variables.
12034         (frame_larger_than_size): Same.
12035         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
12036         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
12037         Initialize.
12038         * doc/invoke.texi (GCC Command Options): Document option arguments.
12039         Explain byte-size arguments and suffixes.
12040         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
12041         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
12042         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
12043         * doc/options.texi (UInteger): Expand.
12044         (Host_Wide_Int, ByteSize): Document new properties.
12045         * final.c (final_start_function_1): Include sizes in an error message.
12046         * function.c (frame_offset_overflow): Same.
12047         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
12048         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
12049         Diagnose unbounded alloca calls only for limits of less than
12050         PTRDIFF_MAX.
12051         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
12052         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
12053         for alloca(0).
12054         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
12055         only for limits of less than PTRDIFF_MAX.
12056         * langhooks-def.h (lhd_handle_option): Change function argument
12057         to HOST_WIDE_INT.
12058         * langhooks.c (lhd_handle_option): Same.
12059         * langhooks.h (handle_option): Same.
12060         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
12061         ByteSize flags.
12062         (var_type, var_type_struct): Same.
12063         (var_set): Handle ByteSize flag.
12064         * optc-gen.awk: Add comments to output to ease debugging.  Make
12065         use of HOST_WIDE_INT where appropriate.
12066         * opts-gen-save.awk:  Use %lx to format unsigned long.
12067         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
12068         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
12069         arguments.  Parse bytes-size suffixes.
12070         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
12071         (enum_value_to_arg): Same.
12072         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
12073         (handle_option): Adjust.
12074         (generate_option): Change function argument to HOST_WIDE_INT.
12075         (cmdline_handle_error): Adjust.
12076         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
12077         (set_option): Change function argument to HOST_WIDE_INT.
12078         (option_enabled): Handle cl_host_wide_int.
12079         (get_option_state): Handle CLVC_SIZE.
12080         (control_warning_option): Same.
12081         * opts.c (common_handle_option): Change function argument to
12082         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
12083         OPT_Wvla_larger_than_.
12084         * opts.h (enum cl_var_type): Add an enumerator.
12085         * stor-layout.c (layout_decl): Print a more meaningful warning.
12086         * toplev.c (output_stack_usage): Adjust.
12088 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
12090         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
12091         call to inline_expand_builtin_string_cmp.
12092         (expand_builtin_strcmp): Likewise.
12093         (expand_builtin_strncmp): Likewise.
12094         (inline_string_cmp): Delete the last parameter, change char_type_node
12095         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
12096         two operands.
12097         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
12098         the inlining expansion on target where the type of the call has same or
12099         narrower precision than unsigned char.
12101 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
12103         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
12104         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
12105         * common.opt (fsave-optimization-record): New option.
12106         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
12107         * doc/invoke.texi (-fsave-optimization-record): New option.
12108         * dumpfile.c: Include "optinfo-emit-json.h".
12109         (struct kv_pair): Move to coretypes.h.
12110         (optgroup_options): Make non-static.
12111         (dump_context::end_scope): Call
12112         optimization_records_maybe_pop_dump_scope.
12113         * dumpfile.h (optgroup_options): New decl.
12114         * json.cc: New file.
12115         * json.h: New file.
12116         * optinfo-emit-json.cc: New file.
12117         * optinfo-emit-json.h: New file.
12118         * optinfo.cc: Include "optinfo-emit-json.h".
12119         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
12120         (optinfo_enabled_p): Check optimization_records_enabled_p.
12121         (optinfo_wants_inlining_info_p): Likewise.
12122         * optinfo.h: Update comment.
12123         * profile-count.c (profile_quality_as_string): New function.
12124         * profile-count.h (profile_quality_as_string): New decl.
12125         (profile_count::quality): New accessor.
12126         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
12127         and optinfo_emit_json_cc_tests.
12128         * selftest.h (selftest::json_cc_tests): New decl.
12129         (selftest::optinfo_emit_json_cc_tests): New decl.
12130         * toplev.c: Include "optinfo-emit-json.h".
12131         (compile_file): Call optimization_records_finish.
12132         (do_compile): Call optimization_records_start.
12133         * tree-ssa-live.c: Include optinfo.h.
12134         (remove_unused_scope_block_p): Retain inlining information if
12135         optinfo_wants_inlining_info_p returns true.
12137 2018-07-20  Richard Biener  <rguenther@suse.de>
12139         PR debug/86585
12140         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
12141         to cover -flto-partition=none.
12143 2018-07-20  Martin Liska  <mliska@suse.cz>
12145         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
12146         (get_decl_source_range): Remove unused function.
12148 2018-07-20  Richard Biener  <rguenther@suse.de>
12150         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
12151         (struct vn_phi_s): Likewise.
12152         (struct vn_reference_s): Likewise.
12153         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
12154         for searching the slot of an entry known to be in the hash itself.
12155         (vn_phi_hasher::equal): Likewise.
12156         (vn_reference_hasher::equal): Likewise.
12157         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
12158         globals.
12159         (optimistic_info, current_info): Remove, keeping only valid_info.
12160         (vn_reference_lookup_1): Remove fallback lookup.
12161         (vn_reference_lookup_2): Likewise.
12162         (vn_nary_op_lookup_1): Likewise.
12163         (vn_phi_lookup): Likewise.
12164         (vn_nary_build_or_lookup_1): Make sure to not chain the built
12165         hash element.
12166         (vn_reference_insert): Adjust, chain the inserted hash element
12167         at last_inserted_ref.
12168         (vn_reference_insert_pieces): Likewise.
12169         (visit_reference_op_call): Likewise.
12170         (vn_nary_op_insert_into): Chain the inserted hash element at
12171         last_inserted_nary.
12172         (vn_nary_op_insert_pieces): Adjust.
12173         (vn_nary_op_insert): Likewise.
12174         (vn_nary_op_insert_stmt): Likewise.
12175         (vn_phi_insert): Adjust, chain the inserted hash element at
12176         last_inserted_phi.
12177         (process_scc): Remove clearing and copying the optimistic
12178         table.  Instead remove elements inserted during an optimistic
12179         iteration from the single table we maintain.
12180         (init_scc_vn): Adjust.
12181         (free_scc_vn): Likewise.
12182         (sccvn_dom_walker::record_cond): Likewise.
12183         (sccvn_dom_walker::after_dom_children): Likewise.
12185 2018-07-19  Martin Sebor  <msebor@redhat.com>
12187         PR tree-optimization/84047
12188         PR tree-optimization/83776
12189         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
12190         (check_array_bounds): Call it.
12192 2018-07-19  Martin Sebor  <msebor@redhat.com>
12194         * align.h (align_flags): Use member initialization.
12196 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
12198         * Makefile.in (OBJS): Add optinfo.o.
12199         * coretypes.h (class symtab_node): New forward decl.
12200         (struct cgraph_node): New forward decl.
12201         (class varpool_node): New forward decl.
12202         * dump-context.h: New file.
12203         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
12204         "tree-pass.h".
12205         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
12206         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
12207         (set_alt_dump_file): Likewise.
12208         (dump_context::~dump_context): New dtor.
12209         (dump_gimple_stmt): Move implementation to...
12210         (dump_context::dump_gimple_stmt): ...this new member function.
12211         Add the stmt to any pending optinfo, creating one if need be.
12212         (dump_gimple_stmt_loc): Move implementation to...
12213         (dump_context::dump_gimple_stmt_loc): ...this new member function.
12214         Start a new optinfo and add the stmt to it.
12215         (dump_gimple_expr): Move implementation to...
12216         (dump_context::dump_gimple_expr): ...this new member function.
12217         Add the stmt to any pending optinfo, creating one if need be.
12218         (dump_gimple_expr_loc): Move implementation to...
12219         (dump_context::dump_gimple_expr_loc): ...this new member function.
12220         Start a new optinfo and add the stmt to it.
12221         (dump_generic_expr): Move implementation to...
12222         (dump_context::dump_generic_expr): ...this new member function.
12223         Add the tree to any pending optinfo, creating one if need be.
12224         (dump_generic_expr_loc): Move implementation to...
12225         (dump_context::dump_generic_expr_loc): ...this new member
12226         function.  Add the tree to any pending optinfo, creating one if
12227         need be.
12228         (dump_printf): Move implementation to...
12229         (dump_context::dump_printf_va): ...this new member function.  Add
12230         the text to any pending optinfo, creating one if need be.
12231         (dump_printf_loc): Move implementation to...
12232         (dump_context::dump_printf_loc_va): ...this new member function.
12233         Start a new optinfo and add the stmt to it.
12234         (dump_dec): Move implementation to...
12235         (dump_context::dump_dec): ...this new member function.  Add the
12236         value to any pending optinfo, creating one if need be.
12237         (dump_context::dump_symtab_node): New member function.
12238         (dump_context::get_scope_depth): New member function.
12239         (dump_context::begin_scope): New member function.
12240         (dump_context::end_scope): New member function.
12241         (dump_context::ensure_pending_optinfo): New member function.
12242         (dump_context::begin_next_optinfo): New member function.
12243         (dump_context::end_any_optinfo): New member function.
12244         (dump_context::s_current): New global.
12245         (dump_context::s_default): New global.
12246         (dump_scope_depth): Delete global.
12247         (dumpfile_ensure_any_optinfo_are_flushed): New function.
12248         (dump_symtab_node): New function.
12249         (get_dump_scope_depth): Reimplement in terms of dump_context.
12250         (dump_begin_scope): Likewise.
12251         (dump_end_scope): Likewise.
12252         (selftest::temp_dump_context::temp_dump_context): New ctor.
12253         (selftest::temp_dump_context::~temp_dump_context): New dtor.
12254         (selftest::verify_item): New function.
12255         (ASSERT_IS_TEXT): New macro.
12256         (ASSERT_IS_TREE): New macro.
12257         (ASSERT_IS_GIMPLE): New macro.
12258         (selftest::test_capture_of_dump_calls): New test.
12259         (selftest::dumpfile_c_tests): Call it.
12260         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
12261         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
12262         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
12263         descriptive comment.
12264         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
12265         (dump_node, dump_bb): Move these unrelated decls.
12266         (class dump_manager): Add leading comment.
12267         * optinfo.cc: New file.
12268         * optinfo.h: New file.
12270 2018-07-19  Michael Collison  <michael.collison@arm.com>
12271             Richard Henderson <rth@redhat.com>
12273         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
12274         (subti): Handle op1 zero.
12275         (subvti4, usub4ti4): New.
12276         (*sub<GPI>3_compare1_imm): New.
12277         (sub<GPI>3_carryinCV): New.
12278         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
12279         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
12281 2018-07-19  Michael Collison  <michael.collison@arm.com>
12282             Richard Henderson <rth@redhat.com>
12284         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
12285         (addti3): Create simpler code if low part is already known to be 0.
12286         (addvti4, uaddvti4): New.
12287         (*add<GPI>3_compareC_cconly_imm): New.
12288         (*add<GPI>3_compareC_cconly): New.
12289         (*add<GPI>3_compareC_imm): New.
12290         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
12291         handle constants within this pattern..
12292         (*add<GPI>3_compareV_cconly_imm): New.
12293         (*add<GPI>3_compareV_cconly): New.
12294         (*add<GPI>3_compareV_imm): New.
12295         (add<GPI>3_compareV): New.
12296         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
12297         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
12298         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
12299         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
12300         with 'comparison' operator.
12301         (*add<GPI>3_compareV_cconly_imm): Ditto.
12302         (*add<GPI>3_compareV_cconly): Ditto.
12303         (*add<GPI>3_compareV_imm): Ditto.
12304         (add<GPI>3_compareV): Ditto.
12305         (add<mode>3_carryinC): Ditto.
12306         (*add<mode>3_carryinC_zero): Ditto.
12307         (*add<mode>3_carryinC): Ditto.
12308         (add<mode>3_carryinV): Ditto.
12309         (*add<mode>3_carryinV_zero): Ditto.
12310         (*add<mode>3_carryinV): Ditto.
12312 2018-07-19  Michael Collison  <michael.collison@arm.com>
12313             Richard Henderson <rth@redhat.com>
12315         * config/aarch64/aarch64-modes.def (CC_V): New.
12316         * config/aarch64/aarch64-protos.h
12317         (aarch64_addti_scratch_regs): Declare
12318         (aarch64_subvti_scratch_regs): Declare.
12319         (aarch64_expand_subvti): Declare.
12320         (aarch64_gen_unlikely_cbranch): Declare
12321         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
12322         for signed overflow using CC_Vmode.
12323         (aarch64_get_condition_code_1): Handle CC_Vmode.
12324         (aarch64_gen_unlikely_cbranch): New function.
12325         (aarch64_addti_scratch_regs): New function.
12326         (aarch64_subvti_scratch_regs): New function.
12327         (aarch64_expand_subvti): New function.
12329 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12331         * config/aarch64/aarch64-option-extensions.def: New entry for profile
12332         extension.
12333         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
12334         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
12335         extension.
12337 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12339         PR target/83009
12340         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
12341         address check not strict.
12343 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12345         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
12346         Umq with Umn.
12347         (store_pair_lanes<mode>): Likewise.
12348         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
12349         enum value 'ADDR_QUERY_LDP_STP_N'.
12350         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
12351         (aarch64_print_address_internal): Add declaration.
12352         (aarch64_print_ldpstp_address): Remove.
12353         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
12354         (aarch64_print_operand): Change printing of 'y'.
12355         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
12356         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
12357         'true' rather than '1'.
12358         * config/aarch64/constraints.md (Uml): Likewise.
12359         (Uml): Rename to Umn.
12360         (Umq): Remove.
12362 2018-07-19  Richard Biener  <rguenther@suse.de>
12364         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
12365         a trailing array.
12366         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
12367         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
12368         (vn_reference_hasher): Likewise.
12369         (struct vn_tables_s): Remove obstack and alloc-pool members.
12370         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
12371         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
12372         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
12373         (vn_reference_insert_pieces): Likewise.
12374         (alloc_vn_nary_op_noinit): Adjust.
12375         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
12376         (vn_phi_eq): Adjust.
12377         (shared_lookup_phiargs): Remove.
12378         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
12379         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
12380         (visit_reference_op_call): Likewise.
12381         (copy_nary, copy_phi, copy_reference): Remove.
12382         (process_scc): Rewind the obstack when iterating.  Do not
12383         copy the elements to valid_info but just move them from one
12384         hashtable to the other.
12385         (allocate_vn_table): Adjust.
12386         (free_vn_table): Likewise.
12387         (init_scc_vn): Likewise.
12388         (free_scc_vn): Likewise.
12390 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
12392         PR target/86560
12393         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
12394         indirect_return as function type attribute.
12395         (ix86_attribute_table): Change indirect_return to function
12396         type attribute.
12397         * doc/extend.texi: Update indirect_return attribute.
12399 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
12401         * wide-int.h (widest2_int): New.
12402         * gimple-fold.c (arith_overflowed_p): Use it.
12403         * tree.h (widest2_int_cst): New.
12404         * tree-vrp.c (wide_int_binop_overflow): Rename from
12405         vrp_int_const_binop.
12406         Rewrite to work on trees.
12407         (extract_range_from_multiplicative_op_1): Abstract code to...
12408         (wide_int_range_min_max): ...here.
12409         (wide_int_range_cross_product): ...and here.
12410         (extract_range_from_binary_expr_1): Abstract overflow code to...
12411         (wide_int_range_mult_wrapping): ...here.
12412         * tree-vrp.h (wide_int_range_cross_product): New.
12413         (wide_int_range_mult_wrapping): New.
12415 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
12416             Julia Koval  <julia.koval@intel.com>
12418         * config/i386/x86-tune-costs.h (skylake_memcpy,
12419         skylake_memset): Replace rep_prefix with unrolling for size 512.
12421 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
12423         PR middle-end/86544
12424         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
12425         comparision with EQ_EXPR in last stmt.
12427 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12429         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
12430         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
12431         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
12432         previously known as "PowerPC AltiVec Built-in Functions".  Move
12433         some material to new subsubsections "PowerPC AltiVec Built-in
12434         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
12435         ISA 2.07".
12436         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
12437         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
12438         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
12439         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
12441 2018-07-18  Richard Biener  <rguenther@suse.de>
12443         PR tree-optimization/86557
12444         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
12445         EXACT_DIV_EXPR.
12447 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
12449         * config/s390/s390.c (s390_function_profiler): Generate CFI.
12451 2018-07-17  Jeff Law  <law@redhat.com>
12453         * config/arm/arm.c (get_label_padding): Update for recent
12454         changes to label_to_alignment.
12456         PR tree-optimization/86010
12457         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
12459         * config/mips/mips.c (vr4130_align_insns): Update for recent
12460         changes to label_to_alignment.
12462         * config/frv/frv.c (frv_label_align): Update for recent changes
12463         to label_to_alignment.
12465         * config/nios2/nios2.c (nios2_label_align): Update for recent
12466         changes which dropped ALIGN_LABELS_LOG.
12468 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
12470         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
12471         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
12473 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
12475         * config/arc/arc.c (arc_label_align): Use align_labels instead of
12476         deprecated align_labels_log.
12478 2018-07-17  Richard Biener  <rguenther@suse.de>
12480         PR lto/86456
12481         * dwarf2out.c (init_sections_and_labels): Always generate
12482         a debug_line_str_section for early LTO debug.
12483         (dwarf2out_finish): Reset debug_line_str_hash output early.
12484         Bump counter for extra dwarf5 .debug_loc labels to not conflict
12485         with fat LTO part.
12486         (dwarf2out_early_finish): Output debug_line_str.
12488 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
12490         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
12491         index register on z196 or later.
12493 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
12495         * config/s390/s390.c (s390_default_align): Set default function
12496         alignment to 16.
12497         (s390_override_options_after_change): Call s390_default align.
12498         (s390_option_override_internal): Call s390_default_align.
12499         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
12501 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
12503         PR middle-end/86542
12504         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
12505         to _looptemp_ clauses, other than the first two.
12507 2018-07-17  Martin Liska  <mliska@suse.cz>
12509         * opts.c: Do not enable OPT_falign_* for -Os.
12511 2018-07-17  Martin Liska  <mliska@suse.cz>
12513         * align.h (MAX_CODE_ALIGN): New.
12514         (MAX_CODE_ALIGN_VALUE): New.
12515         * common/config/i386/i386-common.c (ix86_handle_option):
12516         (MAX_CODE_ALIGN): Moved to align.h.
12517         * final.c (MAX_CODE_ALIGN): Likewise.
12518         * opts.c (parse_and_check_align_values):
12519         (MAX_CODE_ALIGN): Likewise.
12520         (MAX_CODE_ALIGN_VALUE): Likewise.
12522 2018-07-17  Martin Liska  <mliska@suse.cz>
12524         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
12525         in order to fulfil coding style.
12526         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
12527         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12528         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
12529         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
12530         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
12531         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
12532         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
12533         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
12534         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12536 2018-07-17  Martin Liska  <mliska@suse.cz>
12538         * align.h: New file.
12539         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
12540         directly.
12541         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
12542         align_flags of label_to_alignment.
12543         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
12544         align_flags class.
12545         * config/m68k/m68k.c: Do not use removed align_labels_value and
12546         align_loops_value.
12547         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
12548         (LOOP_ALIGN): Likewise.
12549         (LABEL_ALIGN): Likewise.
12550         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
12551         Remove not used macro.
12552         (rs6000_loop_align): Change return type to align_flags.
12553         (rs6000_loop_align_max_skip): Remove.
12554         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
12555         Change return type to align_flags.
12556         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
12557         Remove not used macro.
12558         (rs6000_loop_align):  Change return type to align_flags.
12559         (rs6000_loop_align_max_skip): Remove.
12560         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
12561         * config/rx/rx-protos.h (rx_align_for_label): Make it
12562         static function.
12563         * config/rx/rx.c (rx_align_for_label): Change return type
12564         to align_flags.
12565         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
12566         macro definitions.
12567         into align_flags class.
12568         (LABEL_ALIGN): Likewise.
12569         (LOOP_ALIGN): Likewise.
12570         * config/s390/s390.c (s390_label_align): Use align_flags
12571         class member.
12572         (s390_asm_output_function_label): Likewise.
12573         * config/sh/sh.c (sh_override_options_after_change):
12574         Use align_flags class directly without macros.
12575         (find_barrier): Likewise.
12576         (barrier_align): Likewise.
12577         (sh_loop_align): Likewise.
12578         * config/spu/spu.c (spu_option_override):
12579         Use align_flags_tuple::get_value instead of removed macros.
12580         (spu_sched_init): Likewise.
12581         * config/spu/spu.h (GTY): Likewise.
12582         * config/visium/visium.c (visium_option_override):
12583         Set "8" as default secondary alignment.
12584         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
12585         in order to guarantee secondary alignment of 8.
12586         * coretypes.h: Include align.h header file.
12587         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
12588         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
12589         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
12590         * doc/tm.texi.in: Likewise.
12591         * final.c (struct label_alignment): Remove not used structure.
12592         (LABEL_ALIGN): Change type to align_flags.
12593         (LOOP_ALIGN): Likewise.
12594         (JUMP_ALIGN): Likewise.
12595         (default_loop_align_max_skip): Remove.
12596         (default_label_align_max_skip): Likewise.
12597         (default_jump_align_max_skip): Likewise.
12598         (default_label_align_after_barrier_max_skip):
12599         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
12600         (LABEL_TO_MAX_SKIP): Remove.
12601         (label_to_alignment): Return align_flags type instead of integer.
12602         (label_to_max_skip): Remove.
12603         (align_fuzz): Use align_flags type.
12604         (compute_alignments): Use align_flags type and use align_flags::max
12605         to combine multiple alignments.
12606         (grow_label_align): Grow vec instead of C array.
12607         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
12608         (shorten_branches):  Use align_flags type and use align_flags::max
12609         to combine multiple alignments.
12610         (final_scan_insn_1): Remove usage of secondary alignment that comes
12611         from label alignment, but instead use proper secondary alignment
12612         which is computed in grow_label_align.
12613         * flags.h (struct align_flags_tuple): Move to align.h.
12614         (struct align_flags): Likewise.
12615         (state_align_loops): Rename to align_loops.
12616         (state_align_jumps): Rename to align_jumps.
12617         (state_align_labels): Rename to align_labels.
12618         (state_align_functions): Rename to align_functions.
12619         (align_loops_log): Remove.
12620         (align_jumps_log): Remove.
12621         (align_labels_log): Remove.
12622         (align_functions_log): Remove.
12623         (align_loops_max_skip): Remove.
12624         (align_jumps_max_skip): Remove.
12625         (align_labels_max_skip): Remove.
12626         (align_functions_max_skip): Remove.
12627         (align_loops_value): Remove.
12628         (align_jumps_value): Remove.
12629         (align_labels_value): Remove.
12630         (align_functions_value): Remove.
12631         * output.h (label_to_alignment): Change return type to align_flags.
12632         (label_to_max_skip): Remove.
12633         * target.def: Remove loop_align_max_skip, label_align_max_skip,
12634         jump_align_max_skip macros.
12635         * targhooks.h (default_loop_align_max_skip): Remove.
12636         (default_label_align_max_skip): Likewise.
12637         (default_jump_align_max_skip): Likewise.
12638         (default_label_align_after_barrier_max_skip): Remove.
12639         * toplev.c (read_log_maxskip): Use ::normalize function.
12640         (parse_N_M): Remove not used argument and also call ::normalize.
12641         (parse_alignment_opts): Do not pass unused arguments.
12642         * varasm.c (assemble_start_function): Use directly align_functions
12643         instead of removed macros.
12644         * system.h: Do not poison removed macros.
12646 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
12648         PR middle-end/86539
12649         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
12650         and cond temporaries don't have reference type if iterator has
12651         pointer type.  For init use &for_pre_body instead of pre_p if
12652         for_pre_body is non-empty.
12654 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12656         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
12657         double-double modes to SFmode directly directly.
12658         (trunc<mode>sf2_fprs): Delete.
12660 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12662         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
12663         for conversions between IFmode and the decimal floating point modes.
12664         (init_float128_ieee): Use the correct names for conversions between
12665         KFmode and the decimal floating point modes.
12667 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12669         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
12670         for the conversions between TDmode and IFmode.
12671         (init_float128_ieee): Use more correct names for the conversions
12672         between TDmode and KFmode.
12674 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
12676         PR tree-optimization/86526
12677         * builtins.c (expand_builtin_memcmp): Formatting fixes.
12678         (inline_expand_builtin_string_cmp): Likewise.
12679         (inline_string_cmp): Likewise.  Use c_readstr instead of
12680         builtin_memcpy_read_str.  Add unit_mode temporary.
12682 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12684         PR middle-end/86528
12685         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
12686         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
12688 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12690         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
12691         Alphabetize prototypes of built-in functions, separating out
12692         built-in functions that are listed in this section but should be
12693         described elsewhere.
12695 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
12697         PR target/86511
12698         * expmed.c (emit_store_flag): Do not emit setcc followed by a
12699         conditional move when trapping comparison was split to a
12700         non-trapping one (and vice versa).
12702 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12704         * config/s390/s390.c (s390_function_profiler): Generate nops
12705         instead of profiler call sequences.
12706         * config/s390/s390.opt: Add the new option.
12708 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12710         * config/s390/s390.c (s390_function_profiler): Generate
12711         __mcount_loc section.
12712         * config/s390/s390.opt: Add the new option.
12714 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12716         * common.opt: Add the new warning.
12717         * config/s390/s390.c (s390_function_profiler): Emit "brasl
12718         %r0,__fentry__" when -mfentry is specified.
12719         (s390_option_override_internal): Disallow -mfentry for 31-bit
12720         CPUs.
12721         * config/s390/s390.opt: Add the new option.
12723 2018-07-16  Richard Biener  <rguenther@suse.de>
12725         PR lto/86523
12726         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
12727         for function-local FUNCTION_DECL and RESULT_DECL immediately.
12729 2018-07-16  Martin Liska  <mliska@suse.cz>
12731         PR ipa/86529
12732         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
12733         to ::get_create.
12735 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
12737         * config/arc/arcHS.md: Update ARCHS scheduling rules.
12739 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
12741         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
12742         for ARCHS4x.
12743         * config/arc/arc-cpus.def (hs4x): New cpu.
12744         (hs4xd): Likewise.
12745         * config/arc/arc-tables.opt: Regenerate.
12746         * config/arc/arc.c (arc_sched_issue_rate): New function.
12747         (TARGET_SCHED_ISSUE_RATE): Define.
12748         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
12749         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
12750         fpu_cvt.
12751         (attr tune): Add ARCHS4x tune values.
12752         (attr tune_dspmpy): Define.
12753         (*tst): Correct instruction type.
12754         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
12755         * config/arc/arcHS4x.md: New file.
12756         * config/arc/fpu.md: Update instruction type attributes.
12757         * config/arc/t-multilib: Regenerate.
12759 2018-07-16  Tom de Vries  <tdevries@suse.de>
12761         PR debug/86455
12762         * var-tracking.c (vt_initialize): Fix pre_dec handling.
12764 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
12766         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
12767         early clobber.
12769 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
12771         PR tree-optimization/86514
12772         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
12773         conversion to a boolean type from a type with greater precision.
12775 2018-07-16  Tom de Vries  <tdevries@suse.de>
12777         * var-tracking.c (vt_initialize): Print adjusted insn slim if
12778         dump_flags request TDF_SLIM.
12780 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
12782         * fold-const.c (int_const_binop_1): Abstract...
12783         (wide_int_binop): ...wide int code here.
12784         (poly_int_binop): ...poly int code here.
12785         Abstract the rest of int_const_binop_1 into int_const_binop.
12786         * fold-const.h (wide_int_binop): New.
12787         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
12788         Remove useless PLUS/MINUS_EXPR case.
12789         (zero_nonzero_bits_from_vr): Move wide int code...
12790         (zero_nonzero_bits_from_bounds): ...here.
12791         (extract_range_from_binary_expr_1): Move mask optimization code...
12792         (range_easy_mask_min_max): ...here.
12793         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
12794         (range_easy_mask_min_max): New.
12796 2018-07-15  Jeff Law  <law@redhat.com>
12798         PR target/85993
12799         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
12800         block.
12802 2018-07-14  Jim Wilson  <jimw@sifive.com>
12804         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
12806 2018-07-14  Paul Koning  <ni1d@arrl.net>
12808         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
12810 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
12812         * lto-streamer-out.c (copy_function_or_variable): Dump info about
12813         copying section.
12815 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
12816             Steve Munroe  <munroesj52@gmail.com>
12818         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
12819         (_mm_andnot_si128): Likewise.
12820         (_mm_or_si128): Likewise.
12821         (_mm_xor_si128): Likewise.
12823 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
12825         PR middle-end/78809
12826         * builtins.c (expand_builtin_memcmp): Inline the calls first
12827         when result_eq is false.
12828         (expand_builtin_strcmp): Inline the calls first.
12829         (expand_builtin_strncmp): Likewise.
12830         (inline_string_cmp): New routine. Expand a string compare
12831         call by using a sequence of char comparison.
12832         (inline_expand_builtin_string_cmp): New routine. Inline expansion
12833         a call to str(n)cmp/memcmp.
12834         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
12835         New option.
12836         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
12838 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
12840         * config/arm/driver-arm.c: Include arm-native.h.
12841         (host_detect_local_cpu): Use auto-generated data tables.
12842         (vendors, arm_cpu_table): Delete.  Move part information to ...
12843         * config/arm/arm-cpus.in: ... here.
12844         * config/arm/parsecpu.awk (gen_native): New function.
12845         (vendor, part): New CPU fields.
12846         (END): Add support for building the native CPU detection tables.
12847         * config/arm/t-arm (arm-native.h): Add build rule.
12848         (driver-arm.o): Add dependency on arm-native.h.
12850 2018-07-13  Richard Biener  <rguenther@suse.de>
12852         PR middle-end/85974
12853         * match.pd (addr1 - addr2): Allow either of the operand to
12854         have a conversion.
12856 2018-07-13  Tom de Vries  <tdevries@suse.de>
12858         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
12859         in remap_ssa_name.
12861 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
12863         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
12864         arrays instead of numbered variables.
12866 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
12868         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
12869         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
12871 2018-07-13  Richard Biener  <rguenther@suse.de>
12873         PR debug/86452
12874         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
12875         instead of get_context_die.
12877 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
12878             Richard Biener  <rguenther@suse.de>
12880         PR middle-end/86489
12881         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
12882         that the loop latch destination where phi is defined.
12884 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
12886         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
12887         (riscv_expand_epilogue): Add assertion to check interrupt mode.
12888         (riscv_set_current_function): Extract getting interrupt type to new
12889         function.
12890         (riscv_get_interrupt_type): New function.
12891         (riscv_merge_decl_attributes): New function, checking interrupt type is
12892         same.
12893         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
12895 2018-07-12  Paul Koning  <ni1d@arrl.net>
12897         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
12898         directive.
12900 2018-07-12  Paul Koning  <ni1d@arrl.net>
12902         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
12903         zero reference, add doloop_end instead.
12904         * doc/md.texi (decrement_and_branch_until_zero): Remove.
12905         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
12906         detail for doloop_end.
12908 2018-07-12  Martin Sebor  <msebor@redhat.com>
12910         PR c/86453
12911         * attribs.c (decl_attributes): Reject conflicting attributes before
12912         calling attribute handlers.
12914 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
12916         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
12917          parameter.
12918         (gcc::dump_manager::get_dump_file_name): likewise.
12919         (dump_begin): Likewise.
12920         * dumpfile.h (dump_begin): Update prototype.
12921         (gcc::dump_manager::get_dump_file_name,
12922         gcc::dump_manager::get_dump_file_name): Update prototype.
12924 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12926         * internal-fn.h (vectorizable_internal_fn_p): New function.
12927         * tree-vect-slp.c (compatible_calls_p): Likewise.
12928         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
12929         to internal functions.
12930         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
12932 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12934         * fold-const.h (inverse_conditions_p): Declare.
12935         * fold-const.c (inverse_conditions_p): New function.
12936         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
12937         that test the inverse condition of a conditional internal function.
12938         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
12939         * internal-fn.c (internal_fn_mask_index): Handle conditional
12940         internal functions.
12941         (vectorized_internal_fn_supported_p): New function.
12942         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
12943         (any_pred_load_store): Replace with...
12944         (need_to_predicate): ...this new variable.
12945         (redundant_ssa_names): New variable.
12946         (ifcvt_can_use_mask_load_store): Move initial checks to...
12947         (ifcvt_can_predicate): ...this new function.  Handle tree codes
12948         for which a conditional internal function exists.
12949         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
12950         instead of ifcvt_can_use_mask_load_store.  Update after variable
12951         name change.
12952         (predicate_load_or_store): New function, split out from
12953         predicate_mem_writes.
12954         (check_redundant_cond_expr): New function.
12955         (value_available_p): Likewise.
12956         (predicate_rhs_code): Likewise.
12957         (predicate_mem_writes): Rename to...
12958         (predicate_statements): ...this.  Use predicate_load_or_store
12959         and predicate_rhs_code.
12960         (combine_blocks, tree_if_conversion): Update after above name changes.
12961         (ifcvt_local_dce): Handle redundant_ssa_names.
12962         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
12963         general conditional functions.
12964         * tree-vect-stmts.c (vectorizable_call): Likewise.
12966 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12967             Alan Hayward  <alan.hayward@arm.com>
12968             David Sherwood  <david.sherwood@arm.com>
12970         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
12971         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
12972         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
12973         plus and minus and convert them into IFN_COND_FMA-based sequences.
12974         (convert_mult_to_fma): Handle conditional plus and minus.
12976 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12978         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
12979         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
12980         (cond_fnms_optab): New optabs.
12981         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
12982         internal functions.
12983         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
12984         * internal-fn.h (get_conditional_internal_fn): Declare.
12985         (get_unconditional_internal_fn): Likewise.
12986         * internal-fn.c (cond_ternary_direct): New macro.
12987         (expand_cond_ternary_optab_fn): Likewise.
12988         (direct_cond_ternary_optab_supported_p): Likewise.
12989         (FOR_EACH_COND_FN_PAIR): Likewise.
12990         (get_conditional_internal_fn): New function.
12991         (get_unconditional_internal_fn): Likewise.
12992         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
12993         (gimple_match_op::gimple_match_op): Add a new overload for 5
12994         operands.
12995         (gimple_match_op::set_op): Likewise.
12996         (gimple_resimplify5): Declare.
12997         * genmatch.c (decision_tree::gen): Generate simplifications for
12998         5 operands.
12999         * gimple-match-head.c (gimple_simplify): Define an overload for
13000         5 operands.  Handle calls with 5 arguments in the top-level overload.
13001         (convert_conditional_op): Handle conversions from unconditional
13002         internal functions to conditional ones.
13003         (gimple_resimplify5): New function.
13004         (build_call_internal): Pass a fifth operand.
13005         (maybe_push_res_to_seq): Likewise.
13006         (try_conditional_simplification): Try converting conditional
13007         internal functions to unconditional internal functions.
13008         Handle 3-operand unconditional forms.
13009         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
13010         Define ternary equivalents of the current rules for binary conditional
13011         internal functions.
13012         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
13013         ternary operations.
13014         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
13015         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
13016         (optab): Handle them.
13017         (SVE_COND_FP_TERNARY): New int iterator.
13018         (sve_fmla_op, sve_fmad_op): New int attributes.
13019         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
13020         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
13021         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
13023 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
13025         * target.def (preferred_else_value): New target hook.
13026         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
13027         * doc/tm.texi: Regenerate.
13028         * targhooks.h (default_preferred_else_value): Declare.
13029         * targhooks.c (default_preferred_else_value): New function.
13030         * internal-fn.h (conditional_internal_fn_code): Declare.
13031         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
13032         (get_conditional_internal_fn): Use it.
13033         (conditional_internal_fn_code): New function.
13034         * gimple-match.h (gimple_match_cond): New struct.
13035         (gimple_match_op): Add a cond member function.
13036         (gimple_match_op::gimple_match_op): Update all forms to take a
13037         gimple_match_cond.
13038         * genmatch.c (expr::gen_transform): Use the same condition as res_op
13039         for the suboperation, but don't specify a particular else_value.
13040         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
13041         (visit_nary_op, visit_reference_op_load): Pass
13042         gimple_match_cond::UNCOND to the gimple_match_op constructor.
13043         * gimple-match-head.c: Include tree-eh.h
13044         (convert_conditional_op): New function.
13045         (maybe_resimplify_conditional_op): Likewise.
13046         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
13047         (gimple_resimplify2): Likewise.
13048         (gimple_resimplify3): Likewise.
13049         (gimple_resimplify4): Likewise.
13050         (maybe_push_res_to_seq): Return null for conditional operations.
13051         (try_conditional_simplification): New function.
13052         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
13053         constructor.
13054         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
13055         IFN_COND_* call.
13056         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
13057         function.
13058         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
13060 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
13062         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
13063         DECL_FCONTEXT
13064         (hash_tree): Do not hash DECL_FCONTEXT
13065         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
13066         Do not stream DECL_FCONTEXT.
13067         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
13068         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
13070 2018-07-12  Richard Biener  <rguenther@suse.de>
13072         PR debug/86462
13073         * dwarf2out.c (gen_block_die): Only output blocks when they have
13074         at least one !DECL_IGNORED_P variable.
13076 2018-07-12  Richard Biener  <rguenther@suse.de>
13078         PR target/84829
13079         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
13080         Remove -mieee-fp handling.
13082 2018-07-12  Richard Biener  <rguenther@suse.de>
13084         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
13085         left-over from last patch.
13087 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
13089         PR tree-optimization/86492
13090         * gimple-ssa-store-merging.c
13091         (imm_store_chain_info::coalesce_immediate_stores): Call
13092         check_no_overlap even for the merge_overlapping case.  Formatting fix.
13094 2018-07-12  Richard Biener  <rguenther@suse.de>
13096         PR middle-end/86479
13097         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
13098         move possibly trapping operations into the conditional.
13100 2018-07-12  Richard Biener  <rguenther@suse.de>
13102         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
13103         (vn_lookup_simplify_result): Remove recursion limit applied
13104         here.
13105         (vn_nary_build_or_lookup_1): Adjust.
13106         (try_to_simplify): Likewise.
13107         * gimple-match-head.c (gimple_resimplify1): Instead apply one
13108         here.
13109         (gimple_resimplify2): Likewise.
13110         (gimple_resimplify3): Likewise.
13111         (gimple_resimplify4): Likewise.
13113 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
13115         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
13116         Use __mmask64 type instead of __mmask8 for __M argument.
13117         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
13118         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
13119         __U argument.
13120         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
13121         __mmask16 for __M argument.
13122         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
13123         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
13124         to __mmask16 instead of __mmask8.
13125         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
13126         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
13127         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
13128         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
13129         instead of __mmask16 for __U argument.
13130         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
13131         __mmask16 instead of __mmask8 for __U argument.
13132         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
13133         __U argument.
13134         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
13135         __mmask16.
13136         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
13137         argument.
13138         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
13139         __U argument.
13140         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
13141         __mmask16.
13142         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
13143         of __mmask16.
13144         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
13145         __U argument.
13146         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
13147         __U argument.
13148         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
13149         __U argument.
13150         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
13151         __U argument.
13152         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
13153         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
13154         return type as well as __M argument type and all casts from __mmask8
13155         to __mmask32.
13156         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
13157         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
13158         return type as well as __M argument type and all casts from __mmask8
13159         to __mmask16.
13160         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
13161         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
13162         return type as well as __M argument type and all casts from __mmask8
13163         to __mmask32.
13164         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
13165         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
13166         return type as well as __M argument type and all casts from __mmask8
13167         to __mmask16.
13168         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
13169         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
13170         __mmask16.
13172 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
13174         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
13175         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
13176         for __U argument.
13178 2018-07-11  Paul Koning  <ni1d@arrl.net>
13180         * doc/md.texi (define_subst): Document how multiple occurrences of
13181         the same argument in the replacement pattern are handled.
13183 2018-07-11  Paul Koning  <ni1d@arrl.net>
13185         * doc/extend.texi (Common Variable Attributes): Move "mode" into
13186         alphabetical order.
13187         (Common Type Attributes): Add "mode" attribute.
13189 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
13191         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
13192         stream DECL_ORIGINAL_TYPE.
13193         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
13194         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
13195         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
13196         Do not walk original type.
13197         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
13198         external decls.
13199         (write_ts_decl_non_common_tree_pointers): Do not stream
13200         DECL_ORIGINAL_TYPE
13201         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
13202         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
13204 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
13206         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
13207         thread twice from the same starting edge.
13209 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
13211         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
13212         * gimple.c (gimple_call_nonnull_result_p): ...here...
13213         (gimple_call_nonnull_arg): ...and here.
13214         * gimple.h (gimple_call_nonnull_result_p): New.
13215         (gimple_call_nonnull_arg): New.
13217 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
13219         * config/arm/arm-cpus.in: Move information from fpu field of each
13220         cpu definition to the isa field.
13221         * config/arm/parsecpu.awk (fpu): Delete match rule.
13222         (gen_comm_data): Don't add bits from the CPU's FPU entry.
13224 2018-07-11  Richard Biener  <rguenther@suse.de>
13226         PR debug/86457
13227         * dwarf2out.c (init_sections_and_labels): Use
13228         output_asm_line_debug_info consistently.
13229         (dwarf2out_early_finish): Likewise.
13230         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
13231         type units.
13233 2018-07-11  Richard Biener  <rguenther@suse.de>
13235         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
13236         Rework father_bb setting in a way to avoid propagating constants
13237         multiple times on a loop body.
13239 2018-07-10  Mark Wielaard  <mark@klomp.org>
13241         PR debug/86459
13242         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
13244 2018-07-10  Richard Biener  <rguenther@suse.de>
13246         * hash-map.h (hash_map::iterator::operator*): Return
13247         references to key and value.
13249 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
13251         PR c++/86443
13252         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
13253         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
13254         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
13255         (gimplify_omp_for): For composite loops, move outer
13256         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
13257         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
13258         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
13259         TREE_LIST for both the original class iterator and the "last" helper
13260         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
13261         loop, remember has_decl_expr from outer composite loops for the
13262         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
13264 2018-07-09  Martin Sebor  <msebor@redhat.com>
13266         PR middle-end/77357
13267         PR middle-end/86428
13268         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
13269         accessing implicitly initialized array elements.
13270         * expr.c (string_constant): Handle string initializers of
13271         character arrays within aggregates.
13272         * gimple-fold.c (fold_array_ctor_reference): Add argument.
13273         Store element offset.  As a special case, handle zero size.
13274         (fold_nonarray_ctor_reference): Same.
13275         (fold_ctor_reference): Add argument.  Store subobject offset.
13276         * gimple-fold.h (fold_ctor_reference): Add argument.
13278 2018-07-09  Paul Koning  <ni1d@arrl.net>
13280         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
13281         (pdp11_insn_cost): New function.
13282         (pdp11_md_asm_adjust): New function.
13283         (TARGET_INVALID_WITHIN_DOLOOP): Define.
13284         (pdp11_rtx_costs): Update to match machine better.
13285         (output_addr_const_pdp11): Correct format mismatch warnings.
13286         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
13287         * config/pdp11/pdp11.md: General change to add base_cost and/or
13288         length attributes for use by new pdp11_insn_cost function.
13289         (MIN_BRANCH): Correct definition.
13290         (MIN_SOB): Ditto.
13291         (doloop_end): Use standard pattern name for looping pattern.
13292         (doloop_end_nocc): New.
13293         (movsf): Add another constraint alternative.
13294         (zero_extendqihi2): Add constraint alternatives for not in place
13295         extend.
13296         (zero_extendhisi2): Remove.
13297         (shift patterns): Add CC handling variants.
13298         (bswaphi2): New.
13299         (bswapsi2): New.
13300         (rothi3): New.
13301         (define_peephole2): New peephole to recognize mov that sets CC for
13302         subsequent test.
13304 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13306         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
13307         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
13308         wi::add.
13310 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
13312         PR c/86420
13313         * real.c (real_nextafter): Return true if result is denormal.
13315 2018-07-09  Martin Liska  <mliska@suse.cz>
13317         * common.opt: Add back wrongly removed attribute.
13319 2018-07-09  Richard Biener  <rguenther@suse.de>
13321         PR debug/86413
13322         * dwarf2out.c (gen_block_die): For an early generated DIE
13323         always output high/low PC attributes.
13325 2018-07-09  Tom de Vries  <tdevries@suse.de>
13327         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
13328         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
13329         onto VAR_DECL with abstract origin.
13331 2018-07-07  Jim Wilson  <jimw@sifive.com>
13333         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
13335 2018-07-07  Tom de Vries  <tdevries@suse.de>
13337         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
13339 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
13341         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
13342         overflow_type.
13343         (combine_bound): Use wide-int overflow calculation instead of
13344         rolling our own.
13345         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
13346         overflow_type.
13347         * fold-const.c (int_const_binop_2): Same.
13348         (extract_muldiv_1): Same.
13349         (fold_div_compare): Same.
13350         (fold_abs_const): Same.
13351         * match.pd: Same.
13352         * poly-int.h (add): Same.
13353         (sub): Same.
13354         (neg): Same.
13355         (mul): Same.
13356         * predict.c (predict_iv_comparison): Same.
13357         * profile-count.c (slow_safe_scale_64bit): Same.
13358         * simplify-rtx.c (simplify_const_binary_operation): Same.
13359         * tree-chrec.c (tree_fold_binomial): Same.
13360         * tree-data-ref.c (split_constant_offset_1): Same.
13361         * tree-if-conv.c (idx_within_array_bound): Same.
13362         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
13363         * tree-ssa-phiopt.c (minmax_replacement): Same.
13364         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
13365         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
13366         * vr-values.c (vr_values::adjust_range_with_scev): Same.
13367         * wide-int.cc (wi::add_large): Same.
13368         (wi::mul_internal): Same.
13369         (wi::sub_large): Same.
13370         (wi::divmod_internal): Same.
13371         * wide-int.h: Change overflow type to overflow_type for neg, add,
13372         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
13373         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
13374         mul_internal, divmod_internal.
13375         (overflow_type): New enum.
13376         (accumulate_overflow): New.
13378 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
13380         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
13381         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
13383 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
13385         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
13386         argument is checked for zero before entering loop, avoid checking again.
13388 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
13390         * gimplify.h (generic_expr_could_trap_p): Set as global function.
13391         * gimplify.h (generic_expr_could_trap_p): Likwise.
13392         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
13394 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
13396         PR tree-optimization/86401
13397         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
13398         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
13399         (fold_bit_and_mask): ... here.  New helper function for match.pd.
13400         * fold-const.h (fold_bit_and_mask): Declare.
13401         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
13403 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
13405         PR target/86324
13406         * target.def (translate_mode_attribute): New hook.
13407         * targhooks.h (default_translate_mode_attribute): Declare.
13408         * targhooks.c (default_translate_mode_attribute): New function.
13409         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
13410         * doc/tm.texi: Regenerate.
13411         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
13412         (rs6000_translate_mode_attribute): New function.
13414 2018-07-06  Paul Koning  <ni1d@arrl.net>
13416         * doc/md.texi (define_split): Document DONE and FAIL.
13417         (define_peephole2): Ditto.
13419 2018-07-05  Jeff Law  <law@redhat.com>
13421         PR tree-optimization/86010
13422         * tree-ssa-dse.c (compute_trims): More aggressively trim at
13423         both the head and tail of mem* and str* calls.
13425 2018-07-05  Jim Wilson  <jimw@sifive.com>
13427         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
13429 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
13431         * config/aarch64/aarch64-simd.md: correct flags text for
13432         MIN_EXPR replacement.
13434 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
13436         * configure: Regenerated.
13438 2018-07-05  Carl Love  <cel@us.ibm.com>
13440         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
13441         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
13442         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
13443         VSX_BUILTIN_DOUBLEL_V4SF.
13445 2018-07-05  Martin Sebor  <msebor@redhat.com>
13447         PR c++/86400
13448         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
13449         than its domain to compute its the upper bound of a char array.
13451 2018-07-05  Nathan Sidwell  <nathan@acm.org>
13453         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
13454         * doc/cpp.texi: Update comment.
13455         * doc/tm.texi: Rebuilt.
13456         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
13457         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
13458         * doc/extend.texi (Backwards Compatibility): Clarify it is system
13459         headers affected by extern "C".
13460         * system.h: Poison NO_IMPLICIT_EXTERN_C.
13461         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
13462         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
13463         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
13464         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
13465         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
13466         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
13467         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
13468         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
13469         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
13470         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
13471         config/sparc/sp64-elf.h, config/spu/spu.h,
13472         config/stormy16/stormy16.h, config/v850/v850.h,
13473         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
13474         define NO_IMPLICIT_EXTERN_C.
13475         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
13477 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13479         PR target/84711
13480         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
13481         instead of GET_MODE_SIZE when comparing Units.
13483 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13485         PR target/84711
13486         * rtlanal.c (set_noop_p): Constrain on mode change,
13487         include hard-reg-set.h
13489 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13491         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
13493 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
13495         Revert
13496         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
13498         PR sanitizer/84250
13499         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
13500         libasan.
13501         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
13503 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
13505         PR sanitizer/84250
13506         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
13507         libasan.
13508         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
13510 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
13512         PR middle-end/86380
13513         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
13515 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
13517         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
13518         neg_*_op* variables.
13520 2018-07-04  Martin Liska  <mliska@suse.cz>
13522         * tree-switch-conversion.c: Define
13523         max_ratio_for_speed and max_ratio_for_size constants.
13525 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
13526             Martin Liska  <mliska@suse.cz>
13528         PR middle-end/66240
13529         PR target/45996
13530         PR c/84100
13531         * common.opt: Rename align options with 'str_' prefix.
13532         * common/config/i386/i386-common.c (set_malign_value): New
13533         function.
13534         (ix86_handle_option): Use it to set -falign-* options/
13535         * config/aarch64/aarch64-protos.h (struct tune_params): Change
13536         type from int to string.
13537         * config/aarch64/aarch64.c: Update default values from int
13538         to string.
13539         * config/alpha/alpha.c (alpha_override_options_after_change):
13540         Likewise.
13541         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
13542         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13543         max skip conditionally.
13544         * config/i386/freebsd.h (SUBALIGN_LOG): New.
13545         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13546         max skip conditionally.
13547         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13548         max skip conditionally.
13549         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
13550         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13551         max skip conditionally.
13552         * config/i386/i386.c (struct ptt): Change type from int to
13553         string.
13554         (ix86_default_align): Set default values.
13555         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
13556         max skip conditionally.
13557         * config/i386/iamcu.h (SUBALIGN_LOG): New.
13558         (ASM_OUTPUT_MAX_SKIP_ALIGN):
13559         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
13560         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13561         max skip conditionally.
13562         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
13563         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
13564         * config/i386/x86-64.h (SUBALIGN_LOG): New.
13565         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13566         max skip conditionally.
13567         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
13568         * config/ia64/ia64.c (ia64_option_override): Set default values
13569         for alignment options.
13570         * config/m68k/m68k.c: Handle new str_align_* options.
13571         * config/mips/mips.c (mips_set_compression_mode): Change
13572         type of constants.
13573         (mips_option_override): Set default values for options.
13574         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
13575         Likewise.
13576         * config/rs6000/rs6000.c (rs6000_option_override_internal):
13577         Likewise.
13578         * config/rx/rx.c (rx_option_override): Likewise.
13579         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
13580         (LABEL_ALIGN): Use align_labels_log.
13581         (LOOP_ALIGN): Use align_loops_align.
13582         * config/s390/s390.c (s390_asm_output_function_label): Use new
13583         macros.
13584         * config/sh/sh.c (sh_override_options_after_change):
13585         Change type of constants.
13586         * config/spu/spu.c (spu_sched_init): Likewise.
13587         * config/sparc/sparc.c (sparc_option_override): Set default
13588         values for options.
13589         * config/visium/visium.c (visium_option_override): Likewise.
13590         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
13591         emit p2align format with last argument if it's not needed.
13592         * doc/invoke.texi: Document extended format of -falign-*.
13593         * final.c: Use align_labels alignment.
13594         * flags.h (struct target_flag_state): Change type to use
13595         align_flags.
13596         (struct align_flags_tuple): New.
13597         (struct align_flags): Likewise.
13598         (align_loops_log): Redefine macro to use new types.
13599         (align_loops_max_skip): Redefine macro to use new types.
13600         (align_jumps_log): Redefine macro to use new types.
13601         (align_jumps_max_skip): Redefine macro to use new types.
13602         (align_labels_log): Redefine macro to use new types.
13603         (align_labels_max_skip): Redefine macro to use new types.
13604         (align_functions_log): Redefine macro to use new types.
13605         (align_loops): Redefine macro to use new types.
13606         (align_jumps): Redefine macro to use new types.
13607         (align_labels): Redefine macro to use new types.
13608         (align_functions): Redefine macro to use new types.
13609         (align_functions_max_skip): Redefine macro to use new types.
13610         (align_loops_value): New macro.
13611         (align_jumps_value): New macro.
13612         (align_labels_value): New macro.
13613         (align_functions_value): New macro.
13614         * function.c (invoke_set_current_function_hook): Propagate
13615         alignment values from flags to global variables default in
13616         topleev.h.
13617         * ipa-icf.c (sem_function::equals_wpa): Use
13618         cl_optimization_option_eq instead of memcmp.
13619         * lto-streamer.h (cl_optimization_stream_out): Support streaming
13620         of string types.
13621         (cl_optimization_stream_in): Likewise.
13622         * optc-save-gen.awk: Support strings in cl_optimization.
13623         * opth-gen.awk: Likewise.
13624         * opts.c (finish_options): Remove error checking of invalid
13625         value ranges.
13626         (MAX_CODE_ALIGN): Remove.
13627         (MAX_CODE_ALIGN_VALUE): Likewise.
13628         (parse_and_check_align_values): New function.
13629         (check_alignment_argument): Likewise.
13630         (common_handle_option): Use check_alignment_argument.
13631         * opts.h (parse_and_check_align_values): Declare.
13632         * toplev.c (init_alignments): Remove.
13633         (read_log_maxskip): New.
13634         (parse_N_M): Likewise.
13635         (parse_alignment_opts): Likewise.
13636         (backend_init_target): Remove usage of init_alignments.
13637         * toplev.h (parse_alignment_opts): Declare.
13638         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
13639         argument.
13640         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
13641         * tree.c (cl_option_hasher::equal): New.
13642         * varasm.c: Use new global macros.
13644 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
13646         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
13647         Use a simpler align directive also if MAXSKIP = ALIGN-1.
13648         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13649         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13650         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13651         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
13652         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
13653         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
13654         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
13655         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13656         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13657         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13658         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13660 2018-07-04  Martin Liska  <mliska@suse.cz>
13661             Jonathan Wakely  <jwakely@redhat.com>
13663         * coverage.c: Use correct type.
13664         * doc/invoke.texi: Language correction.
13666 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
13668         PR target/85620
13669         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
13670         ENDBRANCH for non-tail call which may return via indirect branch.
13671         * doc/extend.texi: Document indirect_return attribute.
13673 2018-07-03  Martin Sebor  <msebor@redhat.com>
13675         PR tree-optimization/86274
13676         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
13677         precondition.
13678         (format_floating): Correct handling of infinities and NaNs.
13680 2018-07-03  Martin Sebor  <msebor@redhat.com>
13682         * print-tree.c (print_real_cst): New function.
13683         (print_node_brief): Call it.
13684         (print_node): Ditto.
13686 2018-07-03  Jeff Law  <law@redhat.com>
13688         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
13689         into a single pattern.
13691         * config/h8300/h8300.md (ors code_iterator): New.
13692         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
13693         a single pattern and single splitter.
13694         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
13695         (iorqi3_1, xorqi3_1): Likewise.
13696         (iorqi3, xorqi3 expanders): Similarly.
13698         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
13699         (movmd_internal) into a single pattern using the P mode iterator.
13700         (movmd splitters): Similarly.
13701         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
13702         (movsd splitters): Similarly.
13704         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
13705         ADDB, ADDW and ADDL into a single ADD attribute which selects the
13706         right table based on the size of the operand.
13707         * config/h8300/h8300.md (length_table): Corresponding changes. All
13708         references to "addb", "addw" and "addl" changed to "add".
13709         (btst patterns): Merge two variants into a single pattern.
13710         (tstqi, tsthi): Likewise.
13711         (addhi3_incdec, addsi3_incdec): Likewise.
13712         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
13713         (mulhi3, mulsi3): Likewise.
13714         (udivhi3, udivsi3): Likewise.
13715         (divhi3, divsi3): Likewise.
13716         (andorqi3, andorhi3, andorsi3): Likewise.
13718 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
13720         PR target/85694
13721         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
13722         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
13724 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13726         PR tree-optimization/85694
13727         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
13728         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
13729         UNSPEC_URHADD.
13730         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
13731         (<u>avg<mode>3_ceil): New patterns.
13733 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
13735         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
13736         scan-tree-dump directive.
13737         * gcc.dg/vect/slp-perm-2.c: Likewise.
13738         * gcc.dg/vect/slp-perm-3.c: Likewise.
13739         * gcc.dg/vect/slp-perm-5.c: Likewise.
13740         * gcc.dg/vect/slp-perm-6.c: Likewise.
13741         * gcc.dg/vect/slp-perm-7.c: Likewise.
13742         * gcc.dg/vect/slp-perm-8.c: Likewise.
13744 2018-07-03  Marek Polacek  <polacek@redhat.com>
13746         PR middle-end/86202
13747         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
13749 2018-07-03  Richard Biener  <rguenther@suse.de>
13751         PR ipa/86389
13752         * tree-ssa-structalias.c (find_func_clobbers): Properly
13753         handle indirect calls.
13755 2018-07-03  Jeff Law  <law@redhat.com>
13757         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
13758         (shifts): New code iterator.
13759         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
13760         expander.  Fix HImode handling on H8/SX.
13761         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
13762         (subqi3, subhi3, subsi3 expanders): Likewise.
13763         (andqi3, andhi3, andsi3 expanders): Likewise.
13764         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
13765         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
13766         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
13767         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
13768         (zero_extendqihi2, zero_extendqisi2): Likewise.
13769         (extendqihi2, extendqisi2): Likewise.
13770         (rotlqi3, rotlhi3, rotlsi3): Likewise.
13771         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
13772         (rotlqi3_1, rotlhi3_1): Likewise.
13773         (logicalhi3_sn, logicalsi3_sn): Likewise.
13774         (logicalhi3, logicalsi3): Likewise.
13776 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13778         * tree-vect-patterns.c (vect_recog_rotate_pattern)
13779         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
13780         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
13781         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
13782         type to append_pattern_def_seq instead of creating a stmt_vec_info
13783         directly.
13784         (build_mask_conversion): Likewise.  Remove vinfo argument.
13785         (vect_add_conversion_to_patterm): Likewise, renaming to...
13786         (vect_add_conversion_to_pattern): ...this.
13787         (vect_recog_mask_conversion_pattern): Update call to
13788         build_mask_conversion.  Pass the vector type to
13789         append_pattern_def_seq here too.
13790         (vect_recog_gather_scatter_pattern): Update call to
13791         vect_add_conversion_to_pattern.
13793 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13795         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
13796         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
13797         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
13798         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
13799         STMT_VINFO_PATTERN_DEF_SEQ to null here.
13800         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
13801         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
13802         append_pattern_def_seq instead of new_pattern_def_seq.
13803         (vect_recog_divmod_pattern): Do both of the above.
13804         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
13805         is null.
13807 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13809         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
13810         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
13811         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
13812         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
13813         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
13814         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
13815         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
13816         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
13817         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
13818         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
13819         parameter with a single stmt_vec_info.
13820         (vect_recog_func_ptr): Likewise.
13821         (vect_recog_gather_scatter_pattern): Likewise, folding in...
13822         (vect_try_gather_scatter_pattern): ...this.
13823         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
13824         the stmt_vec_info of the statement to be matched.  Don't clear
13825         STMT_VINFO_RELATED_STMT.
13826         (vect_pattern_recog): Update call accordingly.
13828 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13830         PR tree-optimization/85694
13831         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
13832         (uavgM3_ceil): Document new optabs.
13833         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
13834         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
13835         functions.
13836         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
13837         (savg_ceil_optab): New optabs.
13838         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
13839         (vect_vect_recog_func_ptrs): Add it.
13840         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
13841         constant directly from the associated lhs.
13843 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13845         * tree-vect-patterns.c (vect_split_statement): New function.
13846         (vect_convert_input): Use it to try to split an existing cast.
13848 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13850         * poly-int.h (print_hex): New function.
13851         * dumpfile.h (dump_dec, dump_hex): Declare.
13852         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
13853         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
13854         min_input_precision, operation_precision and operation_sign.
13855         * tree-vect-patterns.c (vect_get_range_info): New function.
13856         (vect_same_loop_or_bb_p, vect_single_imm_use)
13857         (vect_operation_fits_smaller_type): Delete.
13858         (vect_look_through_possible_promotion): Add an optional
13859         single_use_p parameter.
13860         (vect_recog_over_widening_pattern): Rewrite to use new
13861         stmt_vec_info infomration.  Handle one operation at a time.
13862         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
13863         (vect_truncatable_operation_p, vect_set_operation_type)
13864         (vect_set_min_input_precision): New functions.
13865         (vect_determine_min_output_precision_1): Likewise.
13866         (vect_determine_min_output_precision): Likewise.
13867         (vect_determine_precisions_from_range): Likewise.
13868         (vect_determine_precisions_from_users): Likewise.
13869         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
13870         (vect_vect_recog_func_ptrs): Put over_widening first.
13871         Add cast_forwprop.
13872         (vect_pattern_recog): Call vect_determine_precisions.
13874 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13876         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
13877         statements that have been replaced by further pattern statements.
13878         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
13880 2018-07-03  Richard Biener  <rguenther@suse.de>
13882         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
13883         always set *dt.  Dump vectype in vectype overload.
13884         * dumpfile.h (dump_gimple_expr): New function.
13885         (dump_gimple_expr_loc): Likewise.
13886         * dumpfile.c (dump_gimple_expr): New function.
13887         (dump_gimple_expr_loc): Likewise.
13889 2018-07-02  Jeff Law  <law@redhat.com>
13891         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
13892         the H8/300, H8/300H and H8/S variants into a single pattern.
13893         (movhi_h8300, movqi_h8300hs): Similarly.
13894         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
13895         (QHI mode iterator): New.
13897         * config/h8300/h8300.md: Remove trailing whitespace.
13899 2018-07-02  Jim Wilson  <jimw@sifive.com>
13901         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
13902         instead of emit_insn for interrupt returns.
13903         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
13904         (riscv_sret, riscv_uret): Likewise.
13906 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
13908         * pretty-print.c (selftest::test_pp_format): Move save and restore
13909         of quotes to class auto_fix_quotes, and add an instance.
13910         * selftest.c: Include "intl.h".
13911         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
13912         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
13913         * selftest.h (selftest::auto_fix_quotes): New class.
13915 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
13917         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
13918         (aarch64_sve_prepare_conditional_op): Remove.
13919         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
13920         Allow aarch64_simd_reg_or_zero as select operand; remove
13921         the aarch64_sve_prepare_conditional_op call.
13922         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
13923         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
13924         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
13925         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
13926         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
13927         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
13928         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
13929         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
13930         and a splitters to match all of the *_any patterns.
13931         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
13933         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
13934         (SVE_COND_FP_BINARY_REV): Remove.
13935         (sve_int_op_rev, sve_fp_op_rev): New.
13936         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
13937         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
13938         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
13939         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
13940         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
13941         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
13942         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
13943         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
13944         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
13946         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
13947         Remove match_dup 1 from the inner unspec.
13948         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
13950         * config/aarch64/aarch64.md (movprfx): New attr.
13951         (length): Default movprfx to 8.
13952         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
13953         (*madd<SVE_I>, *msub<SVE_I): Likewise.
13954         (*<su>mul<SVE_I>3_highpart): Likewise.
13955         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
13956         (*v<ASHIFT><SVE_I>3): Likewise.
13957         (*<su><MAXMIN><SVE_I>3): Likewise.
13958         (*<su><MAXMIN><SVE_F>3): Likewise.
13959         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
13960         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
13961         (*div<SVE_F>4): Likewise.
13963 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
13965         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
13966         in dump string.
13968 2018-07-02  Richard Biener  <rguenther@suse.de>
13970         PR tree-optimization/86363
13971         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
13972         memset argument refers to a non-variable address.
13974 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
13976         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
13977         {PLUS,MINUS}_EXPR code to...
13978         (adjust_symbolic_bound): ...here,
13979         (combine_bound): ...here,
13980         (set_value_range_with_overflow): ...and here.
13982 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
13984         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
13985         code...
13986         (extract_range_from_abs_expr): ...here.
13988 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
13990         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
13991         -fno-omit-frame-pointer when not optimizing.
13993 2018-07-02  Martin Liska  <mliska@suse.cz>
13995         PR ipa/86279
13996         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
13997         (propagate_nothrow): Likewise.
13999 2018-07-02  Martin Liska  <mliska@suse.cz>
14001         PR ipa/86323
14002         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
14004 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
14006         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
14007         function in r262149, changing "loc" param from source_location to
14008         const dump_location_t &.
14009         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
14010         declaration, as above.
14012 2018-07-01  Paul Koning  <ni1d@arrl.net>
14014         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
14015         -munit-asm, -mgnu-asm, -mdec-asm.
14016         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
14017         (pdp11_output_labelref): New.
14018         (pdp11_output_def): New.
14019         (pdp11_output_addr_vec_elt): New.
14020         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
14021         %# and %@ format codes.
14022         (pdp11_option_override): New.
14023         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
14024         (pdp11_output_ident): New.
14025         (pdp11_asm_named_section): New.
14026         (pdp11_asm_init_sections): New.
14027         (pdp11_file_start): New.
14028         (pdp11_file_end): New.
14029         (output_ascii): Use .ascii/.asciz for -mdec-asm.
14030         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
14031         %o, like %c but octal.
14032         (pdp11_option_override): New.
14033         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
14034         -mdec-asm.
14035         (DATA_SECTION_ASM_OP): Ditto.
14036         (READONLY_DATA_SECTION_ASM_OP): New.
14037         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
14038         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
14039         (ASM_OUTPUT_LABELREF): Ditto.
14040         (ASM_OUTPUT_DEF): Ditto.
14041         (ASM_OUTPUT_EXTERNAL): New.
14042         (ASM_OUTPUT_SOURCE_FILENAME): New.
14043         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
14044         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
14045         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
14046         %# and %@ format codes.
14047         * config/pdp11/pdp11.opt (mgnu-asm): New.
14048         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
14049         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
14050         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
14052 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
14054         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
14055         dereferencing path[] beyond its length.
14056         (debug_path): New.
14057         (debug_all_paths): New.
14058         (rewire_first_differing_edge): New.
14059         (adjust_paths_after_duplication): New.
14060         (duplicate_thread_path): Call adjust_paths_after_duplication.
14061         Add new argument.
14062         (thread_through_all_blocks): Add new argument to
14063         duplicate_thread_path.
14065 2018-06-30  Jim Wilson  <jimw@sifive.com>
14067         * config/riscv/predicates.md (p2m1_shift_operand): New.
14068         (high_mask_shift_operand): New.
14069         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
14070         pattern using p2m1_shift_operand.
14071         (lshsi3_zero_extend_3+2): New combiner pattern using
14072         high_mask_shift_operand.
14074 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14076         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
14077         split out from...
14078         (vect_recog_rotate_pattern): ...here.
14079         (vect_convert_input): Try to insert casts of invariants in the
14080         preheader.
14081         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
14082         preheader to be empty.
14084 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14086         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
14087         vector type.  If given, install it in the new statement's
14088         STMT_VINFO_VECTYPE.
14089         (vect_element_precision): New function.
14090         (vect_unpromoted_value): New struct.
14091         (vect_unpromoted_value::vect_unpromoted_value): New function.
14092         (vect_unpromoted_value::set_op): Likewise.
14093         (vect_look_through_possible_promotion): Likewise.
14094         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
14095         (vect_widened_op_tree, vect_convert_input): Likewise.
14096         (vect_convert_inputs, vect_convert_output): Likewise.
14097         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
14098         to handle the optional cast of the multiplication result and
14099         vect_widened_op_tree to detect the widened multiplication itself.
14100         Do not require the input and output of promotion casts to have
14101         the same sign, but base the signedness of the operation on the
14102         input rather than the result.  If the pattern includes two
14103         promotions, check that those promotions have the same sign.
14104         Do not restrict the MULT_EXPR handling to a double-width result;
14105         handle quadruple-width results and wider.  Use vect_convert_inputs
14106         to convert the inputs to the common type.
14107         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
14108         to handle the optional cast of the ABS result.  Also allow a sign
14109         change or a sign extension between the ABS and MINUS.
14110         Use vect_widened_op_tree to detect the widened subtraction and use
14111         vect_convert_inputs to convert the inputs to the common type.
14112         (vect_handle_widen_op_by_const): Delete.
14113         (vect_recog_widen_op_pattern): New function.
14114         (vect_recog_widen_mult_pattern): Use it.
14115         (vect_recog_widen_shift_pattern): Likewise.
14116         (vect_recog_widen_sum_pattern): Use
14117         vect_look_through_possible_promotion to handle the promoted
14118         PLUS_EXPR operand.
14120 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14122         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
14123         the containing gimple_seq *.
14124         * gimple-iterator.h (gsi_for_stmt): Declare it.
14125         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
14126         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
14127         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
14128         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
14129         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
14130         checks.
14131         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
14132         split out from...
14133         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
14134         statement being replaced is part of an existing pattern
14135         definition sequence, inserting the new pattern statements before
14136         the original one.
14137         (vect_pattern_recog_1): Don't return a bool.  If the statement
14138         is already part of a pattern, instead apply pattern matching
14139         to the pattern definition statements.  Don't clear the
14140         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
14141         (vect_pattern_recog): Don't break after the first match;
14142         continue processing the pattern definition statements instead.
14143         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
14145 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14147         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14148         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
14149         (vect_recog_widen_sum_pattern): Use it.
14151 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14153         * tree-vect-loop.c (vectorizable_reduction): Assert that the
14154         phi is not a pattern statement and has not been replaced by
14155         a pattern statement.
14156         * tree-vect-patterns.c (type_conversion_p): Don't check
14157         STMT_VINFO_IN_PATTERN_P.
14158         (vect_recog_vector_vector_shift_pattern): Likewise.
14159         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
14160         the pattern statement rather than the original statement; check
14161         directly for a WIDEN_MULT_EXPR here.
14162         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
14163         vect_is_simple_use to return the pattern statement rather
14164         than the original statement; use is_pattern_stmt_p to check
14165         for such a pattern statement.
14166         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
14167         to return the pattern statement rather than the original statement;
14168         don't do the same transformation here.
14169         (vect_is_simple_use): If the defining statement has been replaced
14170         by a pattern statement, return the pattern statement instead.
14171         Remove the corresponding (local) transformation from the vectype
14172         overload.
14174 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14176         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
14177         end and default to null.
14178         * tree-vect-loop.c (vect_create_epilog_for_reduction)
14179         (vectorizable_reduction): Update calls accordingly, dropping the
14180         gimple ** argument if the passed-back statement isn't needed.
14181         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
14182         (vect_recog_rotate_pattern): Likewise.
14183         (vect_recog_mask_conversion_pattern): Likewise.
14184         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
14185         (vect_mask_constant_operand_p): Likewise.
14186         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
14187         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
14188         (get_group_load_store_type, get_load_store_type): Likewise.
14189         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
14190         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
14191         (vectorizable_conversion, vectorizable_assignment): Likewise.
14192         (vectorizable_shift, vectorizable_operation): Likewise.
14193         (vectorizable_store, vect_is_simple_cond): Likewise.
14194         (vectorizable_condition, vectorizable_comparison): Likewise.
14195         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
14196         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
14197         and move it to the end.  Cope with null def_stmt_outs.
14199 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14201         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
14203 2018-06-29  Jeff Law  <law@redhat.com>
14205         * config/v850/v850.c (v850_legitimate_address_p): Handle large
14206         displacements for TARGET_V850E2V3 and newer.
14207         (TARGET_LRA_P): Remove.  Defaults to LRA now.
14208         * config/v850/v850.md (sign23byte_load): Remove.
14209         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
14210         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
14212 2018-06-29  Martin Liska  <mliska@suse.cz>
14214         PR lto/85759
14215         * coverage.c (coverage_init): Mangle full path name.
14216         * doc/invoke.texi: Document the change.
14217         * gcov-io.c (mangle_path): New.
14218         * gcov-io.h (mangle_path): Likewise.
14219         * gcov.c (mangle_name): Use mangle_path for path mangling.
14221 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14223         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
14224         if starting source register is not even.
14226 2018-06-29  Martin Liska  <mliska@suse.cz>
14228         PR tree-optimization/86263
14229         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
14230         Make edge redirection.
14232 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
14234         * dumpfile.c (dump_loc): Add indentation based on scope depth.
14235         (dump_scope_depth): New variable.
14236         (get_dump_scope_depth): New function.
14237         (dump_begin_scope): New function.
14238         (dump_end_scope): New function.
14239         * dumpfile.h (get_dump_scope_depth): New declaration.
14240         (dump_begin_scope): New declaration.
14241         (dump_end_scope): New declaration.
14242         (class auto_dump_scope): New class.
14243         (AUTO_DUMP_SCOPE): New macro.
14244         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
14245         AUTO_DUMP_SCOPE.
14247 2018-06-29  Richard Biener  <rguenther@suse.de>
14249         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
14250         compute_all_dependences succeeds.
14251         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
14252         exceed --param loop-max-datarefs-for-datadeps.
14254 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
14256         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
14258 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
14260         PR target/86348
14261         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
14262         alternative 0 in preferred_for_speed attribute.
14264 2018-06-28  Paul Koning  <ni1d@arrl.net>
14266         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
14267         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
14268         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
14269         * config/pdp11/pdp11.md: Correct "length" attribute calculation
14270         for shift insn patterns.
14272 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
14274         * cgraph.c (cgraph_node::get_body): Replace assignments to
14275         "dump_file" with calls to set_dump_file.
14276         * dumpfile.c (alt_dump_file): Make static, and group with...
14277         (alt_flags): ...this definition.
14278         (dumps_are_enabled): New variable.
14279         (refresh_dumps_are_enabled): New function.
14280         (set_dump_file): New function.
14281         (set_alt_dump_file): New function.
14282         (gcc::dump_manager::dump_start): Replace assignments to
14283         "dump_file" and "alt_dump_file" with calls to set_dump_file and
14284         set_alt_dump_file.
14285         (gcc::dump_manager::dump_finish): Likewise.
14286         * dumpfile.h (alt_dump_file): Delete decl.
14287         (dumps_are_enabled): New variable decl.
14288         (set_dump_file): New function decl.
14289         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
14290         global.
14291         * tree-nested.c (lower_nested_functions): Replace assignments to
14292         "dump_file" with calls to set_dump_file.
14294 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
14296         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
14297         goto_locus of each outgoing edge of each basic block.
14299 2018-06-28  Richard Biener  <rguenther@suse.de>
14301         * dwarf2out.c (decl_scope_table): Remove.
14302         (push_decl_scope): Likewise.
14303         (pop_decl_scope): Likewise.
14304         (gen_type_die_for_member): Do not call push/pop_decl_scope.
14305         (gen_struct_or_union_type_die): Likewise.
14306         (gen_tagged_type_die): Likewise.
14307         (dwarf2out_init): Do not initialize decl_scope_table.
14308         (dwarf2out_c_finalize): Do not free it.
14310 2018-06-28  Richard Biener  <rguenther@suse.de>
14312         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
14313         deciding whether to not re-use a DIE.
14315 2018-06-28  Richard Biener  <rguenther@suse.de>
14317         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
14318         DW_AT_abstract_origin attribute.
14320 2018-06-28  Martin Liska  <mliska@suse.cz>
14322         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
14323         Use newly introduced constants.
14324         * tree-switch-conversion.h (struct jump_table_cluster):
14325         Define max_ratio_for_size and max_ratio_for_speed.
14327 2018-06-28  Martin Liska  <mliska@suse.cz>
14329         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
14330         Add new checking assert to catch invalid state.
14331         (jump_table_cluster::can_be_handled): Handle single case
14332         clusters.
14333         (jump_table_cluster::is_beneficial): Bail out for such case.
14334         (bit_test_cluster::find_bit_tests):
14335         Add new checking assert to catch invalid state.
14336         (bit_test_cluster::can_be_handled): Handle single case
14337         clusters.
14338         (bit_test_cluster::is_beneficial): Bail out for such case.
14339         (switch_decision_tree::analyze_switch_statement):
14340         Fix comment.
14342 2018-06-28  Martin Liska  <mliska@suse.cz>
14344         * common.opt: Introduce -completion option.
14345         * gcc.c (driver_handle_option): Handle it.
14346         (driver::main): Print completions if completion
14347         is set.
14348         * opt-suggestions.c (option_proposer::get_completions):
14349         New function.
14350         (option_proposer::suggest_completion): Likewise.
14351         (option_proposer::find_param_completions): Likewise.
14352         (verify_autocompletions): Likewise.
14353         (test_completion_valid_options): Likewise.
14354         (test_completion_valid_params): Likewise.
14355         (in_completion_p): Likewise.
14356         (empty_completion_p): Likewise.
14357         (test_completion_partial_match): Likewise.
14358         (test_completion_garbage): Likewise.
14359         (opt_proposer_c_tests): Likewise.
14360         * opt-suggestions.h: Declare new functions.
14361         * opts.c (common_handle_option): Handle OPT__completion_.
14362         * selftest-run-tests.c (selftest::run_tests): Add
14363         opt_proposer_c_tests.
14364         * selftest.c (assert_str_startswith): New.
14365         * selftest.h (assert_str_startswith): Likewise.
14366         (opt_proposer_c_tests): New.
14367         (ASSERT_STR_STARTSWITH): Likewise.
14369 2018-06-28  Martin Liska  <mliska@suse.cz>
14371         * Makefile.in: Add opt-suggestions.o.
14372         * gcc-main.c: Include opt-suggestions.h.
14373         * gcc.c (driver::driver): Likewise.
14374         (driver::~driver): Remove m_option_suggestions.
14375         (driver::build_option_suggestions): Moved to option_proposer.
14376         (driver::suggest_option): Likewise.
14377         (driver::handle_unrecognized_options): Use option_proposer.
14378         * gcc.h (class driver): Add new memver m_option_proposer.
14379         * opt-suggestions.c: New file.
14380         * opt-suggestions.h: New file.
14382 2018-06-28  Martin Liska  <mliska@suse.cz>
14384         * vec.h (class auto_string_vec): New (moved from auto_argvec).
14385         (auto_string_vec::~auto_string_vec): Likewise.
14387 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
14389         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
14390         prevent_decl_creation_for_types fields up and add reset_location field.
14391         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
14392         statement if id->reset_location is true.
14393         (copy_edges_for_bb): Do not set goto_locus on the new edges if
14394         id->reset_location is true.
14395         (copy_phis_for_bb): Force input_location on the arguments if
14396         id->reset_location is true.
14397         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
14398         is set on the function to be inlined.
14400 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
14402         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
14404 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
14406         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
14407         registers for Pmode.
14408         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
14409         hard registers for the clobbered pseudo.
14411 2018-06-27  Paul Koning  <ni1d@arrl.net>
14413         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
14414         mutually exclusive options.
14415         * config/pdp11/constraints.md (h): New constraint.
14416         (O): Update definition to match shift code generation.
14417         (D): New constraint.
14418         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
14419         (CCFP): Remove.
14420         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
14421         function.
14422         (output_jump): Change arguments.
14423         (pdp11_fixed_cc_regs): New function.
14424         (pdp11_cc_mode): Ditto.
14425         (pdp11_expand_shift): Ditto.
14426         (pdp11_assemble_shift): Ditto.
14427         (pdp11_small_shift): Ditto.
14428         (pdp11_branch_cost): Remove.
14429         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
14430         from output.
14431         (pdp11_register_move_cost): Update for CC registers.
14432         (pdp11_rtx_costs): Add case for LSHIFTRT.
14433         (pdp11_output_jump): Add CCNZ mode conditional branches.
14434         (notice_update_cc_on_set): Remove.
14435         (pdp11_cc_mode): New function.
14436         (simple_memory_operand): Correct pre/post decrement case.
14437         (no_side_effect_operand): New function.
14438         (pdp11_regno_reg_class): Add CC_REGS class.
14439         (pdp11_fixed_cc_regs): New function.
14440         (pdp11_small_shift): New function.
14441         (pdp11_expand_shift): New function to expand shift insns.
14442         (pdp11_assemble_shift): New function to output shifts.
14443         (pdp11_branch_cost): Remove.
14444         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
14445         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
14446         (WCHAR_TYPE): Ditto.
14447         (PTRDIFF_TYPE): Ditto.
14448         (ADJUST_INSN_LENGTH): New macro.
14449         (FIXED_REGISTERS): Add CC registers.
14450         (CALL_USED_REGISTERS): Ditto.
14451         (reg_class): Ditto.
14452         (REG_CLASS_NAMES): Ditto.
14453         (REG_CLASS_CONTENTS): Ditto.
14454         (SELECT_CC_MODE): Use new function.
14455         (TARGET_FLAGS_REGNUM): New macro.
14456         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
14457         (cc0_reg_rtx): Remove.
14458         (CC_STATUS_MDEP): Remove.
14459         (CC_STATUS_MDEFP_INIT): Remove.
14460         (CC_IN_FPU): Remove.
14461         (NOTICE_UPDATE_CC): Remove.
14462         (REGISTER_NAMES): Add CC registers.
14463         (BRANCH_COST): Change to constant 1.
14464         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
14465         handling.
14466         * config/pdp11/pdp11.opt (mbcopy): Remove.
14467         (mbcopy-builtin): Remove.
14468         (mbranch-cheap): Remove.
14469         (mbranch-expensive): Remove.
14470         * config/pdp11/predicates.md (expand_shift_operand): Update to
14471         match shift code generation.
14472         (ccnz_operator): New predicate.
14473         * doc/invoke.texi (PDP-11 Options): Remove deleted options
14474         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
14475         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
14476         exclusive options.
14477         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
14478         description of O constraint.
14480 2018-06-27  Jeff Law  <law@redhat.com>
14481             Austin Law  <austinklaw@gmail.com>
14483         * config/v850/v850.md (addsi3_set_flags): New pattern.
14484         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
14485         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
14486         (zero_extendhisi2_v850_set_flags): Likewise.
14487         (zero_extendqisi2_v850_set_flags): Likewise.
14488         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
14489         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
14490         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
14492         * config/v850/v850-protos.h (notice_update_cc): Remove.
14493         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
14494         (v850_print_operand): Handle 'D' and "d".
14495         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
14496         Add handling of arithmetic/logical operations compared against zero.
14497         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
14498         Do not look at v850_compare_op, instead get mode from last argument.
14499         (v850_gen_compare): Remove
14500         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
14501         after reload for prologue insns.
14502         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
14503         patterns.
14504         (construct_save_jarl): Likewise.
14505         (TARGET_FLAGS_REGNUM): Define.
14506         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
14507         (NOTICE_UPDATE_CC): Remove.
14508         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
14509         than cc0.  Conditionalize on reload_completed.
14510         (cmpsi_insn, setfcc_insn): Likewise.
14511         (tst1 splitter): Turn into define_and_split which sets the flags
14512         after reload.
14513         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
14514         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
14515         (cstoresf4, cstoredf4): Clobber the flags.
14516         (cmpsi, cmpsf, cmpdf): Remove expanders.
14517         (setf_insn): Remove pattern.
14518         (addsi3): Turn into define_and_split which clobbers the flags after
14519         reload and a suitable pattern (addsi3_clobber_flags) for use after
14520         reload.
14521         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
14522         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
14523         (ashrsi3, ashrsi3_v850e2): Likewise.
14524         (bins): Clobber the flags.
14525         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
14526         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
14527         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
14528         (call_value_internal_short, call_value_internal_long): Likewise.
14529         (callt_save_interrupt, callt_return_interrupt): Likewise.
14530         (save_interrupt, return_interrupt): Likewise.
14531         (callt_save_all_interrupt, save_all_interrupt): Likewise.
14532         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
14533         (restore_all_interrupt, _restore_all_interrupt): Likewise.
14534         (All FP comparisons): Only allow after reload has completed.
14535         (trfsr): Likewise.
14536         (divh, divhu): Tweak output template.
14537         (branch_z_normal, branch_z_invert): Remove
14538         (branch_nz_normal, branch_nz_invert): Likewise.
14539         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
14541         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
14542         * config/v850/v850.c (notice_update_cc): Remove.
14543         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
14544         (CC_NO_CARRY): Likewise.
14545         (NOTICE_UPDATE_CC): Define to nothing.
14546         * config/v850/v850.md: Remove block comment on cc0 handling
14547         Remove "cc" attribute from all patterns.  Remove cc_status handling
14548         from all patterns.  Minor formatting fixes.
14550 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14552         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
14553         (cortex-a76.cortex-a55): Likewise.
14554         * config/aarch64/aarch64-tune.md: Regenerate.
14555         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
14556         cortex-a76.cortex-a55.
14558 2018-06-27  Jeff Law  <law@redhat.com>
14560         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
14561         (MULTILIB_DIRNAMES): Similarly.
14563 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
14565         * gimple.h (gimple_return_retbnd): Delete.
14566         (gimple_return_set_retbnd): Likewise.
14567         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
14568         gimple_return_set_retbnd.
14569         * gimple-pretty-print.c (dump_gimple_return): Remove call to
14570         gimple_return_retbnd and adjust.
14571         * tree-inline.h (struct copy_body_data): Remove retbnd field.
14572         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
14573         Explicitly return NULL in a couple more cases.  Move assertion
14574         on debug statements and remove unreachable code.
14575         (reset_debug_binding): Do not test id->retbnd.
14576         (expand_call_inline): Do not set it.
14578 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
14580         * configure.ac: Add --disable-gcov option.
14581         * configure: Regenerate.
14582         * Makefile.in: Honour @enable_gcov@.
14583         * doc/install.texi: Document --disable-gcov.
14585 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14587         * config/arm/arm-cpus.in (cortex-a76): New entry.
14588         (cortex-a76.cortex-a55): Likewise.
14589         * config/arm/arm-tables.opt: Regenerate.
14590         * config/arm/arm-tune.md: Likewise.
14591         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
14592         * doc/invoke.texi (ARM Options): Document cortex-a76 and
14593         cortex-a76.cortex-a55.
14595 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
14597         PR target/85769
14598         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
14600 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
14602         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
14603         comment.
14604         (EPILOGUE_USES): Likewise.
14606 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
14608         * tree-inline.c (remap_location): New function extracted from...
14609         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
14610         (copy_phis_for_bb): ...here.  Call remap_location.
14611         (copy_cfg_body): Adjust call to copy_edges_for_bb.
14613 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
14615         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
14616         unaligned vsx for 16B memset.
14618 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
14620         PR target/86285
14621         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
14622         ieee128_float_type_node to long_double_type_node unless
14623         TARGET_LONG_DOUBLE_128 is set.
14625 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
14627         * cfgloop.c (get_loop_location): Convert return type from
14628         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
14629         by implicit construction from rtx_insn *, and using
14630         dump_user_location_t::from_function_decl for the fallback case.
14631         * cfgloop.h (get_loop_location): Convert return type from
14632         location_t to dump_user_location_t.
14633         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
14634         dump_printf_loc to pass in a dump_location_t rather than a
14635         location_t, via the gimple stmt.
14636         * coverage.c (get_coverage_counts): Update calls to
14637         dump_printf_loc to pass in dump_location_t rather than a
14638         location_t.
14639         * doc/optinfo.texi (Dump types): Convert example of
14640         dump_printf_loc from taking "locus" to taking "insn".  Update
14641         description of the "_loc" calls to cover dump_location_t.
14642         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
14643         "selftest.h".
14644         (dump_user_location_t::dump_user_location_t): New constructors,
14645         from gimple *stmt and rtx_insn *.
14646         (dump_user_location_t::from_function_decl): New function.
14647         (dump_loc): Make static.
14648         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
14649         const dump_location_t &.
14650         (dump_generic_expr_loc): Delete.
14651         (dump_printf_loc): Convert param "loc" from location_t to
14652         const dump_location_t &.
14653         (selftest::test_impl_location): New function.
14654         (selftest::dumpfile_c_tests): New function.
14655         * dumpfile.h: Include "profile-count.h".
14656         (class dump_user_location_t): New class.
14657         (struct dump_impl_location_t): New struct.
14658         (class dump_location_t): New class.
14659         (dump_printf_loc): Convert 2nd param from source_location to
14660         const dump_location_t &.
14661         (dump_generic_expr_loc): Delete.
14662         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
14663         const dump_location_t &.
14664         * gimple-fold.c (fold_gimple_assign): Update call to
14665         dump_printf_loc to pass in a dump_location_t rather than a
14666         location_t, via the gimple stmt.
14667         (gimple_fold_call): Likewise.
14668         * gimple-loop-interchange.cc
14669         (loop_cand::analyze_iloop_reduction_var): Update for change to
14670         check_reduction_path.
14671         (tree_loop_interchange::interchange): Update for change to
14672         find_loop_location.
14673         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
14674         change in return-type of find_loop_location.
14675         (graphite_regenerate_ast_isl): Likewise.
14676         * graphite-optimize-isl.c (optimize_isl): Likewise.
14677         * graphite.c (graphite_transform_loops): Likewise.
14678         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
14679         pass in a dump_location_t rather than a location_t, via the
14680         gimple stmt.
14681         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
14682         * ipa.c (walk_polymorphic_call_targets): Likewise.
14683         * loop-unroll.c (report_unroll): Convert "locus" param from
14684         location_t to dump_location_t.
14685         (decide_unrolling): Update for change to get_loop_location's
14686         return type.
14687         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
14688         location_t to dump_user_location_t.
14689         (grid_find_single_omp_among_assignments_1): Updates calls to
14690         dump_printf_loc to pass in a dump_location_t rather than a
14691         location_t, via the gimple stmt.
14692         (grid_parallel_clauses_gridifiable): Convert "tloc" from
14693         location_t to dump_location_t.  Updates calls to dump_printf_loc
14694         to pass in a dump_location_t rather than a location_t, via the
14695         gimple stmt.
14696         (grid_inner_loop_gridifiable_p): Likewise.
14697         (grid_dist_follows_simple_pattern): Likewise.
14698         (grid_gfor_follows_tiling_pattern): Likewise.
14699         (grid_target_follows_gridifiable_pattern): Likewise.
14700         (grid_attempt_target_gridification): Convert initialization
14701         of local "grid" from memset to zero-initialization; FIXME: does
14702         this require C++11?  Update call to dump_printf_loc to pass in a
14703         optinfo_location rather than a location_t, via the gimple stmt.
14704         * profile.c (read_profile_edge_counts): Updates call to
14705         dump_printf_loc to pass in a dump_location_t rather than a
14706         location_t
14707         (compute_branch_probabilities): Likewise.
14708         * selftest-run-tests.c (selftest::run_tests): Call
14709         dumpfile_c_tests.
14710         * selftest.h (dumpfile_c_tests): New decl.
14711         * tree-loop-distribution.c (pass_loop_distribution::execute):
14712         Update for change in return type of find_loop_location.
14713         * tree-parloops.c (parallelize_loops): Likewise.
14714         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
14715         "locus" from location_t to dump_user_location_t.
14716         (canonicalize_loop_induction_variables): Likewise.
14717         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
14718         for change in return type of find_loop_location.
14719         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
14720         to dump_printf_loc to pass in a dump_location_t rather than a
14721         location_t, via the stmt.
14722         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
14723         Likewise.
14724         * tree-vect-loop-manip.c (find_loop_location): Convert return
14725         type from source_location to dump_user_location_t.
14726         (vect_do_peeling): Update for above change.
14727         (vect_loop_versioning): Update for change in type of
14728         vect_location.
14729         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
14730         from location_t to dump_user_location_t.
14731         (vect_estimate_min_profitable_iters): Update for change in type
14732         of vect_location.
14733         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
14734         location_t to dump_location_t.
14735         (vect_slp_bb): Update for change in type of vect_location.
14736         * tree-vectorizer.c (vect_location): Convert from source_location
14737         to dump_user_location_t.
14738         (try_vectorize_loop_1): Update for change in vect_location's type.
14739         (vectorize_loops): Likewise.
14740         (increase_alignment): Likewise.
14741         * tree-vectorizer.h (vect_location): Convert from source_location
14742         to dump_user_location_t.
14743         (find_loop_location): Convert return type from source_location to
14744         dump_user_location_t.
14745         (check_reduction_path): Convert 1st param from location_t to
14746         dump_user_location_t.
14747         * value-prof.c (check_counter): Update call to dump_printf_loc to
14748         pass in a dump_user_location_t rather than a location_t; update
14749         call to error_at for change in type of "locus".
14750         (check_ic_target): Update call to dump_printf_loc to
14751         pass in a dump_user_location_t rather than a location_t, via the
14752         call_stmt.
14754 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
14756         * config/s390/s390.h (enum processor_flags): Do not use
14757         default tune parameter when -march was specified.
14759 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
14761         PR target/86314
14762         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
14763         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
14764         operands.
14766 2018-06-26  Richard Biener  <rguenther@suse.de>
14768         PR tree-optimization/86287
14769         PR bootstrap/86316
14770         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
14771         (vect_analyze_loop): Initialize n_stmts.
14773 2018-06-26  Richard Biener  <rguenther@suse.de>
14775         PR middle-end/86271
14776         * fold-const.c (fold_convertible_p): Pointer extension
14777         isn't valid.
14779 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
14781         PR debug/86064
14782         * dwarf2out.c (loc_list_has_views): Adjust comments.
14783         (dw_loc_list): Split single cross-partition range with
14784         nonzero locview.
14786 2018-06-25  Jeff Law  <law@redhat.com>
14788         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
14789         on -mbig-switch by default.
14791         * config/v850/predicates.md (const_float_1_operand): Fix match_code
14792         test.
14793         (const_float_0_operand): Remove unused predicate.
14794         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
14795         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
14796         (recipsf2): New expander.  Original pattern now called
14797         (recipsf2_insn).
14798         (recipdf2, recipdf2_insn): Similarly.
14799         (rsqrtsf2, rsqrtsf2_insn): Similarly
14800         (rsqrtdf2, rsqrtdf2_insn): Similarly
14802 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
14804         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
14805         Simplify logic for FreeBSD (twice).
14807 2018-06-25  Martin Sebor  <msebor@redhat.com>
14809         PR tree-optimization/86204
14810         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
14811         a strnlen result if it's less than the length of the string.
14813 2018-06-25  Martin Sebor  <msebor@redhat.com>
14815         PR tree-optimization/85700
14816         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
14817         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
14818         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
14820 2018-06-25  Martin Sebor  <msebor@redhat.com>
14822         * doc/extend.texi (Zero-length arrays): Update and clarify.
14824 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
14826         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
14827         added IEEE/IBM long double multilib support on PowerPC little
14828         endian Linux systems.
14829         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
14830         (MULTILIB_DEFAULTS): Likewise.
14831         * config/rs6000/rs6000.c (rs6000_option_override_internal):
14832         Likewise.
14833         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
14834         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
14835         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
14837 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
14839         PR middle-end/86311
14840         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
14841         (REORDER_45): Likewise.
14843 2018-06-25  Jeff Law  <law@redhat.com>
14845         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
14846         dividend to 32 bits.  Adjust length.
14847         (udivmodhi4): Cleanup output template.  Fix length.
14849 2018-06-25  Carl Love  <cel@us.ibm.com>
14851         * config/rs6000/vsx.md: Change word selector to prefered location.
14853 2018-06-25  Richard Biener  <rguenther@suse.de>
14855         PR tree-optimization/86304
14856         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
14857         epilogue-if-converted loops as well.
14859 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
14861         * lto-section-out.c (lto_begin_section): Do not print section
14862         name for noaddr and unnumbered dumps.
14864 2018-06-25  Richard Biener  <rguenther@suse.de>
14866         * tree-vectorizer.h (struct vec_info_shared): New structure
14867         with parts split out from struct vec_info and loop_nest from
14868         struct _loop_vec_info.
14869         (struct vec_info): Adjust accordingly.
14870         (struct _loop_vec_info): Likewise.
14871         (LOOP_VINFO_LOOP_NEST): Adjust.
14872         (LOOP_VINFO_DATAREFS): Likewise.
14873         (LOOP_VINFO_DDRS): Likewise.
14874         (struct _bb_vec_info): Likewise.
14875         (BB_VINFO_DATAREFS): Likewise.
14876         (BB_VINFO_DDRS): Likewise.
14877         (struct _stmt_vec_info): Add dr_aux member.
14878         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
14879         (DR_MISALIGNMENT_UNINITIALIZED): New.
14880         (set_dr_misalignment): Adjust.
14881         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
14882         (vect_analyze_loop): Adjust prototype.
14883         (vect_analyze_loop_form): Likewise.
14884         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
14885         Compute dependences lazily.
14886         (vect_record_base_alignments): Use shared datarefs/ddrs.
14887         (vect_verify_datarefs_alignment): Likewise.
14888         (vect_analyze_data_refs_alignment): Likewise.
14889         (vect_analyze_data_ref_accesses): Likewise.
14890         (vect_analyze_data_refs): Likewise.
14891         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
14892         constructor parameter for shared part.
14893         (vect_analyze_loop_form): Pass in shared part and adjust.
14894         (vect_analyze_loop_2): Pass in storage for the number of
14895         stmts.  Move loop nest finding to the caller.  Compute
14896         datarefs lazily.
14897         (vect_analyze_loop): Pass in shared part.
14898         (vect_transform_loop): Verify shared datarefs are unchanged.
14899         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
14900         constructor parameter for shared part.
14901         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
14902         (vect_slp_bb): Verify shared datarefs are unchanged before
14903         transform.
14904         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
14905         change.
14906         (new_stmt_vec_info): Initialize DR_AUX misalignment to
14907         DR_MISALIGNMENT_UNINITIALIZED.
14908         * tree-vectorizer.c (vec_info::vec_info): Add constructor
14909         parameter for shared part.
14910         (vec_info::~vec_info): Adjust.
14911         (vec_info_shared::vec_info_shared): New.
14912         (vec_info_shared::~vec_info_shared): Likewise.
14913         (vec_info_shared::save_datarefs): Likewise.
14914         (vec_info_shared::check_datarefs): Likewise.
14915         (try_vectorize_loop_1): Construct shared part live for analyses
14916         of a single loop for multiple vector sizes.
14917         * tree-parloops.c (gather_scalar_reductions): Adjust.
14919 2018-06-25  Richard Biener  <rguenther@suse.de>
14921         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
14922         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
14923         (vect_analyze_data_refs): Remove similar code from here and
14924         simplify accordingly.
14926 2018-06-25  Richard Biener  <rguenther@suse.de>
14928         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
14929         for reverse storage order accesses rather than asserting
14930         they cannot happen here.
14932 2018-06-25  Tom de Vries  <tdevries@suse.de>
14934         PR debug/86257
14935         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
14936         Use data16 instead of .byte for insn prefix.
14938 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
14940         PR C++/86082
14941         * parser.c (make_char_string_pack): Pass this literal chars
14942         through cpp_interpret_string.
14943         (cp_parser_userdef_numeric_literal): Check the result of
14944         make_char_string_pack.
14946 2018-06-24  Maya Rashish  <coypu@sdf.org>
14948         * ginclude/stddef.h: Simplify conditions around avoiding
14949         re-definition of __size_t.
14951 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
14953         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
14954         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
14956 2018-06-22  Maya Rashish  <coypu@sdf.org>
14958         * doc/invoke.texi (mno-fancy-math-387): Update for changes
14959         made to OpenBSD and NetBSD through the years.
14961 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14963         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
14964         behavior of vec_pack (vector double, vector double) to match
14965         behavior of vec_float2 (vector double, vector double).
14967 2018-06-22  Olivier Hainque  <hainque@adacore.com>
14969         * gimplify.c (gimplify_function_tree): Prevent creation
14970         of a trampoline for the address of the current function
14971         passed to entry/exit instrumentation hooks.
14973 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
14975         PR target/86222
14976         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
14977         correctly.
14979 2018-06-22  Martin Liska  <mliska@suse.cz>
14981         PR tree-optimization/86263
14982         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
14983         Bail out if is_enabled is false.
14984         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
14985         New declaration.
14986         (jump_table_cluster::is_enabled): New function.
14988 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
14990         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
14991         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
14992         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
14993         (lto_input_ts_binfo_tree_pointers): Likewise.
14994         * tree-streamer-out.c (streamer_write_tree_bitfields,
14995         write_ts_binfo_tree_pointers): Likewise.
14996         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
14998 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
15000         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
15002 2018-06-22  Martin Liska  <mliska@suse.cz>
15004         * symbol-summary.h (get): Make it pure and inline move
15005         functionality from ::get function.
15006         (get): Remove and inline into ::get and ::get_create.
15007         (get_create): Move code from ::get function.
15009 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
15011         PR target/85994
15012         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
15013         -x assembler-with-cpp.
15015 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
15017         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
15018         _FILE_OFFSET_BITS=64 for C++.
15020 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
15022         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
15023         conversion insn that shows up when pr85657-3.c is compiled using
15024         IEEE 128-bit long double.
15025         (neg<mode>2_internal): Use the correct mode to check whether the
15026         mode is IBM extended.
15027         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
15028         multiply and divide external functions from being created more
15029         than once.
15031 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
15033         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
15034         functions.
15035         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
15036         the edge can be forwarded.
15037         (cfg_layout_merge_blocks): Likewise.
15039 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
15041         * except.c (finish_eh_generation): Commit edge insertions only after
15042         the EH edges have been redirected from post-landing to landing pads.
15044 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
15046         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
15047         create_tmp_var_for to create the FRAME decl.
15048         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
15050 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
15052         * tree-inline.c (copy_edges_for_bb): Minor tweak.
15053         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
15054         debug statement when resetting its value.
15055         (expand_call_inline): Copy the locus of the call onto the assignment
15056         of the return value, if any.  Use local variable in more cases.
15058 2018-06-21  Martin Liska  <mliska@suse.cz>
15060         * ipa-pure-const.c (propagate_nothrow): Use
15061         funct_state_summaries->get.
15062         (dump_malloc_lattice): Likewise.
15063         (propagate_malloc): Likewise.
15065 2018-06-21  Richard Biener  <rguenther@suse.de>
15067         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
15068         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
15069         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
15070         comment.
15071         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
15072         BLOCK_ABSTRACT_ORIGIN unconditionally.
15074 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
15076         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
15077         deleting it.
15078         * ipa-reference.c (ipa_reference_c_finalize): Delete
15079         ipa_ref_opt_sum_summaries and set it to NULL.
15081 2018-06-21  Tom de Vries  <tdevries@suse.de>
15083         PR tree-optimization/85859
15084         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
15085         test with comment from bb_no_side_effects_p.
15087 2018-06-21  Richard Biener  <rguenther@suse.de>
15089         PR tree-optimization/86232
15090         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
15091         max for constant niter.
15093 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15095         * config/aarch64/aarch64-simd.md
15096         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
15098 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15100         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
15101         Make opernads of the unspec commutative.
15103 2018-06-21  Richard Biener  <rguenther@suse.de>
15105         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
15106         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
15107         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
15108         (vect_analyze_data_ref_dependence): Re-order checks to deal with
15109         NULL DR_STEP.
15110         (vect_record_base_alignments): Do not record base alignment
15111         for gathers or scatters.
15112         (vect_compute_data_ref_alignment): Drop return value that is always
15113         true.  Bail out early for gathers or scatters.
15114         (vect_enhance_data_refs_alignment): Bail out early for gathers
15115         or scatters.
15116         (vect_find_same_alignment_drs): Likewise.
15117         (vect_analyze_data_refs_alignment): Remove dead code.
15118         (vect_slp_analyze_and_verify_node_alignment): Likewise.
15119         (vect_analyze_data_refs): For possible gathers or scatters do
15120         not create an alternate DR, just check their possible validity
15121         and mark them.  Adjust DECL_NONALIASED handling to not rely
15122         on DR_BASE_ADDRESS.
15123         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
15124         update inits of gathers or scatters.
15125         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
15126         Also copy gather/scatter flag to pattern vinfo.
15128 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15130         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
15131         behavior of vec_packsu (vector unsigned long long, vector unsigned
15132         long long) to match behavior of vec_packs with same signature.
15134 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
15135             Thomas Schwinge <thomas@codesourcery.com>
15136             Cesar Philippidis  <cesar@codesourcery.com>
15138         * gimplify.c (gimplify_scan_omp_clauses): Add support for
15139         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
15140         (gimplify_adjust_omp_clauses): Likewise.
15141         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
15142         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
15143         (gimplify_omp_target_update): Update handling of acc update and
15144         enter/exit data.
15145         * omp-low.c (install_var_field): Remove unused parameter
15146         base_pointers_restrict.
15147         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
15148         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
15149         FINALIZE}
15150         (omp_target_base_pointers_restrict_p): Delete.
15151         (scan_omp_target): Update call to scan_sharing_clauses.
15152         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
15153         FINALIZE}.
15154         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
15155         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
15156         (convert_local_omp_clauses): Likewise.
15157         * tree-pretty-print.c (dump_omp_clause): Likewise.
15158         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
15159         FINALIZE}.
15160         (omp_clause_code_name): Likewise.
15162 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
15164         PR debug/86194
15165         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
15166         be narrowed.
15168         PR tree-optimization/86231
15169         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
15170         anti-range don't overwrite *vr0min before using it to compute *vr0max.
15172 2018-06-20  Tom de Vries  <tdevries@suse.de>
15174         PR tree-optimization/86097
15175         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
15176         iv type if signedness of iv type is not the same as that of *nit.
15178 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
15180         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
15181         EDGE_EH edges, verify they are all EDGE_EH.
15183 2018-06-20  Maya Rashish  <coypu@sdf.org>
15185         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
15187 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15189         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
15190         * config/aarch64/aarch64.c (xgene1_tunings): Add
15191         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
15192         (aarch64_mode_valid_for_sched_fusion_p):
15193         Allow 16-byte modes.
15194         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
15195         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
15196         128-bit modes.
15197         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
15198         New pattern.
15199         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
15200         * config/aarch64/iterators.md (VQ2): New mode iterator.
15202 2018-06-20  Martin Liska  <mliska@suse.cz>
15204         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
15205         Change default ratio from 10 to 8.
15207 2018-06-20  Martin Liska  <mliska@suse.cz>
15209         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
15210         New.
15211         (bit_test_cluster::find_bit_tests): Likewise.
15212         (switch_decision_tree::analyze_switch_statement): Find clusters.
15213         * tree-switch-conversion.h (struct jump_table_cluster): Document
15214         hierarchy.
15216 2018-06-20  Martin Liska  <mliska@suse.cz>
15218         * tree-switch-conversion.c (switch_conversion::collect):
15219         Record m_uniq property.
15220         (switch_conversion::expand): Bail out for special conditions.
15221         (group_cluster::~group_cluster): New.
15222         (group_cluster::group_cluster): Likewise.
15223         (group_cluster::dump): Likewise.
15224         (jump_table_cluster::emit): New.
15225         (switch_decision_tree::fix_phi_operands_for_edges): New.
15226         (struct case_node): Remove struct.
15227         (jump_table_cluster::can_be_handled): New.
15228         (case_values_threshold): Moved to header.
15229         (reset_out_edges_aux): Likewise.
15230         (jump_table_cluster::is_beneficial): New.
15231         (bit_test_cluster::can_be_handled): Likewise.
15232         (add_case_node): Remove.
15233         (bit_test_cluster::is_beneficial): New.
15234         (case_bit_test::cmp): New.
15235         (bit_test_cluster::emit): New.
15236         (expand_switch_as_decision_tree_p): Remove.
15237         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
15238         (fix_phi_operands_for_edge): Likewise.
15239         (switch_decision_tree::analyze_switch_statement): New.
15240         (compute_cases_per_edge): Move ...
15241         (switch_decision_tree::compute_cases_per_edge): ... here.
15242         (try_switch_expansion): Likewise.
15243         (switch_decision_tree::try_switch_expansion): Likewise.
15244         (record_phi_operand_mapping): Likewise.
15245         (switch_decision_tree::record_phi_operand_mapping): Likewise.
15246         (emit_case_decision_tree): Likewise.
15247         (switch_decision_tree::emit): Likewise.
15248         (balance_case_nodes): Likewise.
15249         (switch_decision_tree::balance_case_nodes): Likewise.
15250         (dump_case_nodes): Likewise.
15251         (switch_decision_tree::dump_case_nodes): Likewise.
15252         (emit_jump): Likewise.
15253         (switch_decision_tree::emit_jump): Likewise.
15254         (emit_cmp_and_jump_insns): Likewise.
15255         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
15256         (emit_case_nodes): Likewise.
15257         (switch_decision_tree::emit_case_nodes): Likewise.
15258         (conditional_probability): Remove.
15259         * tree-switch-conversion.h (enum cluster_type): New.
15260         (PRINT_CASE): New.
15261         (struct cluster): Likewise.
15262         (cluster::cluster): Likewise.
15263         (struct simple_cluster): Likewise.
15264         (simple_cluster::simple_cluster): Likewise.
15265         (struct group_cluster): Likewise.
15266         (struct jump_table_cluster): Likewise.
15267         (struct bit_test_cluster): Likewise.
15268         (struct min_cluster_item): Likewise.
15269         (struct case_tree_node): Likewise.
15270         (case_tree_node::case_tree_node): Likewise.
15271         (jump_table_cluster::case_values_threshold): Likewise.
15272         (struct case_bit_test): Likewise.
15273         (struct switch_decision_tree): Likewise.
15274         (struct switch_conversion): Likewise.
15275         (switch_decision_tree::reset_out_edges_aux): Likewise.
15277 2018-06-20  Martin Liska  <mliska@suse.cz>
15279         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
15280         (hoist_edge_and_branch_if_true): Likewise.
15281         (expand_switch_using_bit_tests_p): Likewise.
15282         (struct case_bit_test): Likewise.
15283         (case_bit_test_cmp): Likewise.
15284         (emit_case_bit_tests): Likewise.
15285         (switch_conversion::switch_conversion): New class.
15286         (struct switch_conv_info): Remove old struct.
15287         (collect_switch_conv_info): More to ...
15288         (switch_conversion::collect): ... this.
15289         (check_range): Likewise.
15290         (switch_conversion::check_range): Likewise.
15291         (check_all_empty_except_final): Likewise.
15292         (switch_conversion::check_all_empty_except_final): Likewise.
15293         (check_final_bb): Likewise.
15294         (switch_conversion::check_final_bb): Likewise.
15295         (create_temp_arrays): Likewise.
15296         (switch_conversion::create_temp_arrays): Likewise.
15297         (free_temp_arrays): Likewise.
15298         (gather_default_values): Likewise.
15299         (switch_conversion::gather_default_values): Likewise.
15300         (build_constructors): Likewise.
15301         (switch_conversion::build_constructors): Likewise.
15302         (constructor_contains_same_values_p): Likewise.
15303         (switch_conversion::contains_same_values_p): Likewise.
15304         (array_value_type): Likewise.
15305         (switch_conversion::array_value_type): Likewise.
15306         (build_one_array): Likewise.
15307         (switch_conversion::build_one_array): Likewise.
15308         (build_arrays): Likewise.
15309         (switch_conversion::build_arrays): Likewise.
15310         (gen_def_assigns): Likewise.
15311         (switch_conversion::gen_def_assigns): Likewise.
15312         (prune_bbs): Likewise.
15313         (switch_conversion::prune_bbs): Likewise.
15314         (fix_phi_nodes): Likewise.
15315         (switch_conversion::fix_phi_nodes): Likewise.
15316         (gen_inbound_check): Likewise.
15317         (switch_conversion::gen_inbound_check): Likewise.
15318         (process_switch): Use the newly created class.
15319         (switch_conversion::expand): New.
15320         (switch_conversion::~switch_conversion): New.
15321         * tree-switch-conversion.h: New file.
15323 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15325         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
15326         tree-vect-patterns.c.
15327         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
15328         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
15329         (vect_recog_sad_pattern): Likewise.
15330         (vect_recog_widen_sum_pattern): Likewise.
15331         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
15332         (vect_recog_widen_shift_pattern): Remove the type_in argument.
15333         (vect_recog_rotate_pattern): Likewise.
15334         (vect_recog_mult_pattern): Likewise.
15335         (vect_recog_vector_vector_shift_pattern): Likewise.
15336         (vect_recog_divmod_pattern): Likewise.
15337         (vect_recog_mixed_size_cond_pattern): Likewise.
15338         (vect_recog_bool_pattern): Likewise.
15339         (vect_recog_mask_conversion_pattern): Likewise.
15340         (vect_try_gather_scatter_pattern): Likewise.
15341         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
15342         (vect_recog_over_widening_pattern): Likewise.
15343         (vect_recog_gather_scatter_pattern): Likewise.
15344         (vect_recog_func_ptr): Move from tree-vectorizer.h
15345         (vect_vect_recog_func_ptrs): Move further down the file.
15346         (vect_recog_func): Likewise.  Remove the third argument.
15347         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
15348         (vect_pattern_recog_1): Expect the pattern function to do any
15349         necessary target tests.  Also expect it to provide a vector type.
15350         Remove the type_in handling.
15352 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15354         * tree-vect-patterns.c (vect_pattern_detected): New function.
15355         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
15356         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
15357         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
15358         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
15359         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
15360         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
15361         (vect_recog_mask_conversion_pattern)
15362         (vect_try_gather_scatter_pattern): Likewise.
15364 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15366         * tree-vect-patterns.c (vect_get_internal_def): New function.
15367         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
15368         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
15369         (search_type_for_mask_1): Use it.
15371 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15373         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
15374         redundant WIDEN_SUM_EXPR handling.
15375         (vect_recog_sad_pattern): Likewise.
15377 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15379         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
15380         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
15381         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
15382         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
15383         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
15385 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15387         * tree-vect-stmts.c (vectorizable_call): Make sure that we
15388         use the stmt_vec_info of the original bb statement for the
15389         new zero assignment, even if the call is part of a pattern.
15391 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15393         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
15394         that the sequence is attached to the original statement rather
15395         than the pattern statement.
15396         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
15397         PATTERN_DEF_SEQ from the original statement rather than
15398         the main pattern statement.
15399         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
15400         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
15401         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
15403 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15405         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
15406         definition statements before the early exit for statements that aren't
15407         live or relevant.
15408         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
15409         split out from...
15410         (vect_transform_loop): ...here.  Process pattern definition
15411         statements without first checking whether the main pattern
15412         statement is live or relevant.
15414 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
15416         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
15417         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
15419 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
15421         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
15422         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
15423         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
15424         (expand_block_compare): Change select_block_compare_mode call.
15425         (expand_strncmp_align_check): Use new functions, fix comment.
15426         (emit_final_str_compare_gpr): New function.
15427         (expand_strn_compare): Refactor and clean up code.
15428         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
15430 2018-06-19  Tony Reix  <tony.reix@atos.com>
15431             Damien Bergamini  <damien.bergamini@atos.com>
15432             David Edelsohn  <dje.gcc@gmail.com>
15434         * collect2.c (static_obj): New variable.
15435         (static_libs): New variable.
15436         (is_in_list): Uncomment declaration.
15437         (main): Track AIX libraries linked statically.
15438         (is_in_list): Uncomment definition.
15439         (scan_prog_file): Don't add AIX shared libraries initializer
15440         to constructor list if linking statically.
15442 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
15444         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
15445         constant.
15446         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
15448 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
15450         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
15451         blocks.
15453 2018-06-19  Martin Liska  <mliska@suse.cz>
15455         * config/i386/i386.c (ix86_can_inline_p): Do not use
15456         ipa_fn_summaries::get_create.
15457         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
15458         get.
15459         (devirtualization_time_bonus): Likewise.
15460         (ipcp_propagate_stage): Likewise.
15461         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
15462         (edge_set_predicate): Likewise.
15463         (evaluate_conditions_for_known_args): Likewise.
15464         (evaluate_properties_for_edge): Likewise.
15465         (ipa_call_summary::reset): Tranform to ...
15466         (ipa_call_summary::~ipa_call_summary): ... this.
15467         (ipa_fn_summary::reset): Transform to ...
15468         (ipa_fn_summary::~ipa_fn_summary): ... this.
15469         (ipa_fn_summary_t::remove): Rename to ...
15470         (ipa_fn_summary_t::remove_callees): ... this.
15471         (ipa_fn_summary_t::duplicate): Use placement new
15472         instead of memory copy.
15473         (ipa_call_summary_t::duplicate): Likewise.
15474         (ipa_call_summary_t::remove): Remove.
15475         (dump_ipa_call_summary): Change get_create to get.
15476         (ipa_dump_fn_summary): Dump only when summary exists.
15477         (analyze_function_body): Use symbol_summary::get instead
15478         of get_create.
15479         (compute_fn_summary): Likewise.
15480         (estimate_edge_devirt_benefit): Likewise.
15481         (estimate_edge_size_and_time): Likewise.
15482         (inline_update_callee_summaries): Likewise.
15483         (remap_edge_change_prob): Likewise.
15484         (remap_edge_summaries): Likewise.
15485         (ipa_merge_fn_summary_after_inlining): Likewise.
15486         (write_ipa_call_summary): Likewise.
15487         (ipa_fn_summary_write): Likewise.
15488         (ipa_free_fn_summary): Likewise.
15489         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
15490         (struct ipa_call_summary): Likewise.
15491         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
15492         of get_create.
15493         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
15494         (estimate_size_after_inlining): Likewise.
15495         (estimate_growth): Likewise.
15496         (growth_likely_positive): Likewise.
15497         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
15498         (inline_call): Likewise.
15499         * ipa-inline.c (caller_growth_limits): Likewise.
15500         (can_inline_edge_p): Likewise.
15501         (can_inline_edge_by_limits_p): Likewise.
15502         (compute_uninlined_call_time): Likewise.
15503         (compute_inlined_call_time): Likewise.
15504         (want_inline_small_function_p): Likewise.
15505         (edge_badness): Likewise.
15506         (update_caller_keys): Likewise.
15507         (update_callee_keys): Likewise.
15508         (inline_small_functions): Likewise.
15509         (inline_to_all_callers_1): Likewise.
15510         (dump_overall_stats): Likewise.
15511         (early_inline_small_functions): Likewise.
15512         (early_inliner): Likewise.
15513         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
15514         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
15515         * ipa-pure-const.c (malloc_candidate_p): Likewise.
15516         * ipa-split.c (execute_split_functions): Likewise.
15517         * symbol-summary.h: Likewise.
15518         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
15520 2018-06-19  Richard Biener  <rguenther@suse.de>
15522         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
15523         (vectorize_loops): ... here.  Fix dbgcnt handling.
15524         (try_vectorize_loop): Wrap try_vectorize_loop_1.
15526 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
15528         PR target/86197
15529         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
15530         ieee128 argument takes up only one (vector) register, not two (floating
15531         point) registers.
15533 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
15535         * gimplify.c (gimplify_init_constructor): Really never clear for an
15536         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
15538 2018-06-19  Richard Biener  <rguenther@suse.de>
15540         PR tree-optimization/86179
15541         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
15542         after failed recognition.
15544 2018-06-18  Martin Sebor  <msebor@redhat.com>
15546         PR middle-end/85602
15547         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
15548         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
15549         Handle integer subtraction.
15550         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
15551         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
15553 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
15555         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
15556         param from rtx to rtx_insn *.
15557         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
15558         param.
15559         (frv_ifcvt_modify_insn): Likwise.
15560         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
15561         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
15562         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
15563         as_a <rtx_insn *> cast to local "unprotected_region" once
15564         it's been established that it's not NULL or pc_rtx.
15565         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
15566         param "sethi" from rtx to rtx_insn *.
15567         (nds32_group_float_insns): Likewise for param "insn".
15568         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
15569         param.
15570         (vax_output_int_subtract): Likewise.
15571         * config/vax/vax.c (vax_output_int_add): Likewise for param
15572         "insn".
15573         (vax_output_int_subtract): Likewise.
15574         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
15575         (emit_pattern_after): Likewise for param "after".
15576         (emit_insn_after): Likewise.
15577         (emit_jump_insn_after): Likewise.
15578         (emit_call_insn_after): Likewise.
15579         (emit_debug_insn_after): Likewise.
15580         (emit_pattern_before): Likewise for param "before".
15581         (emit_insn_before): Likewise.
15582         (emit_jump_insn_before): Likewise.
15583         * final.c (get_insn_template): Likewise for param "insn", removing
15584         a cast.
15585         * output.h (get_insn_template): Likewise for 2nd param.
15586         * rtl.h (emit_insn_before): Likewise.
15587         (emit_jump_insn_before): Likewise.
15588         (emit_debug_insn_before_noloc): Likewise.
15589         (emit_insn_after): Likewise.
15590         (emit_jump_insn_after): Likewise.
15591         (emit_call_insn_after): Likewise.
15592         (emit_debug_insn_after): Likewise.
15593         (set_insn_deleted): Likewise for param.
15595 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
15597         PR target/85358
15598         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
15599         floating point modes, so that IFmode is numerically greater than
15600         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
15601         to declare the ordering.  This prevents IFmode from being
15602         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
15603         machine.  Include rs6000-modes.h to share the fractional values
15604         between genmodes* and the rest of the compiler.
15605         (IFmode): Likewise.
15606         (KFmode): Likewise.
15607         (TFmode): Likewise.
15608         * config/rs6000/rs6000-modes.h: New file.
15609         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
15610         meaning of rs6000_long_double_size so that 126..128 selects an
15611         appropriate 128-bit floating point type.
15612         (rs6000_option_override_internal): Likewise.
15613         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
15614         (TARGET_LONG_DOUBLE_128): Change the meaning of
15615         rs6000_long_double_size so that 126..128 selects an appropriate
15616         128-bit floating point type.
15617         (LONG_DOUBLE_TYPE_SIZE): Update comment.
15618         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
15619         source and destination to match the standard usage.
15620         (truncifkf2): Likewise.
15621         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
15622         ISA 2.07 to use an explicit clobber, instead of passing in a
15623         temporary.
15624         (copysign<mode>3_soft): Likewise.
15626 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
15628         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
15629         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
15630         (vect_slp_analyze_instance_dependence): Likewise.
15631         (vect_enhance_data_refs_alignment): Likewise.
15632         (vect_analyze_data_refs_alignment): Likewise.
15633         (vect_slp_analyze_and_verify_instance_alignment
15634         (vect_analyze_data_ref_accesses): Likewise.
15635         (vect_prune_runtime_alias_test_list): Likewise.
15636         (vect_analyze_data_refs): Likewise.
15637         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
15638         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15639         (vect_analyze_scalar_cycles_1): Likewise.
15640         (vect_get_loop_niters): Likewise.
15641         (vect_analyze_loop_form_1): Likewise.
15642         (vect_update_vf_for_slp): Likewise.
15643         (vect_analyze_loop_operations): Likewise.
15644         (vect_analyze_loop): Likewise.
15645         (vectorizable_induction): Likewise.
15646         (vect_transform_loop): Likewise.
15647         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
15648         * tree-vect-slp.c (vect_analyze_slp): Likewise.
15649         (vect_make_slp_decision): Likewise.
15650         (vect_detect_hybrid_slp): Likewise.
15651         (vect_slp_analyze_operations): Likewise.
15652         (vect_slp_bb): Likewise.
15653         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
15654         (vectorizable_bswap): Likewise.
15655         (vectorizable_call): Likewise.
15656         (vectorizable_simd_clone_call): Likewise.
15657         (vectorizable_conversion): Likewise.
15658         (vectorizable_assignment): Likewise.
15659         (vectorizable_shift): Likewise.
15660         (vectorizable_operation): Likewise.
15661         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
15663 2018-06-18  Martin Sebor  <msebor@redhat.com>
15665         PR tree-optimization/81384
15666         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
15667         * builtins.c (expand_builtin_strnlen): New function.
15668         (expand_builtin): Call it.
15669         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
15670         * builtins.def (BUILT_IN_STRNLEN): New.
15671         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
15672         Warn for bounds in excess of maximum object size.
15673         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
15674         single-value ranges.  Handle strnlen.
15675         (handle_builtin_strlen): Handle strnlen.
15676         (strlen_check_and_optimize_stmt): Same.
15677         * doc/extend.texi (Other Builtins): Document strnlen.
15679 2018-06-18  Maya Rashish  <coypu@sdf.org>
15681         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
15682         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
15683         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
15685         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
15686         here to ...
15687         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
15689 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
15691         * tree.c (escaped_string::escape): Replace cast to char * by
15692         const_cast<char *> (unescaped).
15694 2018-06-18  Nick Clifton  <nickc@redhat.com>
15696         PR 84195
15697         * tree.c (escaped_string): New class.  Converts an unescaped
15698         string into its escaped equivalent.
15699         (warn_deprecated_use): Use the new class to convert the
15700         deprecation message, if present.
15701         (test_escaped_strings): New self test.
15702         (test_c_tests): Add test_escaped_strings.
15703         * doc/extend.texi (deprecated): Add a note that the
15704         deprecation message is affected by the -fmessage-length
15705         option, and that control characters will be escaped.
15706         (#pragma GCC error): Document this pragma.
15707         (#pragma GCC warning): Likewise.
15708         * doc/invoke.texi (-fmessage-length): Document this option's
15709         effect on the #warning and #error preprocessor directives and
15710         the deprecated attribute.
15712 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
15714         * tree.c (decl_value_expr_lookup): Revert latest change.
15715         (decl_value_expr_insert): Likewise.
15717 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
15719         * gimplify.c (nonlocal_vlas): Delete.
15720         (nonlocal_vla_vars): Likewise.
15721         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
15722         referenced VLAs.
15723         (gimplify_body): Do not create and destroy nonlocal_vlas.
15724         * tree-nested.c: Include diagnostic.h.
15725         (use_pointer_in_frame): Tweak.
15726         (lookup_field_for_decl): Add assertion and declare the transformation.
15727         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
15728         internal error when the reference is in a wrong context.  Do not
15729         create a debug decl by default.
15730         (note_nonlocal_block_vlas): Delete.
15731         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
15732         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
15733         create a debug decl by default.
15734         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
15735         call is in a wrong context.
15736         (fixup_vla_decls): New function.
15737         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
15738         debug variables were created.
15739         * tree.c (decl_value_expr_lookup): Add checking assertion.
15740         (decl_value_expr_insert): Likewise.
15742 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
15744         PR middle-end/82479
15745         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
15746         * tree-scalar-evolution.c (interpret_expr): Likewise.
15747         (expression_expensive_p): Likewise.
15748         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
15749         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
15750         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
15751         (ssa_defined_by_minus_one_stmt_p): New.
15753 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
15755         PR middle-end/64946
15756         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
15757         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
15758         * dojump.c (do_jump): Likewise.
15759         * expr.c (expand_expr_real_2): Check operand type's sign.
15760         * fold-const.c (const_unop): Handle ABSU_EXPR.
15761         (fold_abs_const): Likewise.
15762         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
15763         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
15764         (strip_sign_op_1): Likesise.
15765         * match.pd: Add new pattern to generate ABSU_EXPR.
15766         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
15767         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
15768         * tree-eh.c (operation_could_trap_helper_p): Likewise.
15769         * tree-inline.c (estimate_operator_cost): Likewise.
15770         * tree-pretty-print.c (dump_generic_node): Likewise.
15771         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
15772         * tree.def (ABSU_EXPR): New.
15774 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
15776         PR middle-end/86095
15777         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
15778         documented as preserved for backward compatibility only.
15779         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
15781         PR rtl-optimization/86108
15782         * bb-reorder.c (create_forwarder_block): Renamed to ...
15783         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
15784         jump from new landing pad to the second part.
15785         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
15786         Adjust callers.
15788 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
15790         PR middle-end/85878
15791         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
15792         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
15793         Only call store_expr for halves if the mode is the same.
15795         PR middle-end/86123
15796         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
15797         Fix up comment formatting.
15799 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15801         * typed-splay-tree.h (typed_splay_tree::remove): New function.
15802         (typed_splay_tree::closure,
15803         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
15804         (typed_splay_tree::typed_splay_tree,
15805         typed_splay_tree::operator =): Declared private.
15806         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
15807         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
15808         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
15809         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
15810         typed_splay_tree::splay_tree_splay,
15811         typed_splay_tree::splay_tree_foreach_helper,
15812         typed_splay_tree::splay_tree_insert,
15813         typed_splay_tree::splay_tree_remove,
15814         typed_splay_tree::splay_tree_lookup,
15815         typed_splay_tree::splay_tree_predecessor,
15816         typed_splay_tree::splay_tree_successor,
15817         typed_splay_tree::splay_tree_min,
15818         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
15819         (typed_splay_tree::root, typed_splay_tree::comp,
15820         typed_splay_tree::delete_key,
15821         typed_splay_tree::delete_value): New data members.
15822         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
15823         typed_splay_tree::remove.
15825 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
15827         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
15828         -mginv and -mno-ginv to the assembler.
15829         * config/mips/mips.opt (-mcrc): New option.
15830         (-mginv): Likewise.
15831         * doc/invoke.text (-mcrc): Document.
15832         (-mginv): Likewise.
15834 2018-06-15  Nick Clifton  <nickc@redhat.com>
15836         PR 84195
15837         * tree.c (escaped_string): New class.  Converts an unescaped
15838         string into its escaped equivalent.
15839         (warn_deprecated_use): Use the new class to convert the
15840         deprecation message, if present.
15841         (test_escaped_strings): New self test.
15842         (test_c_tests): Add test_escaped_strings.
15843         * doc/extend.texi (deprecated): Add a note that the
15844         deprecation message is affected by the -fmessage-length
15845         option, and that control characters will be escaped.
15846         (#pragma GCC error): Document this pragma.
15847         (#pragma GCC warning): Likewise.
15848         * doc/invoke.texi (-fmessage-length): Document this option's
15849         effect on the #warning and #error preprocessor directives and
15850         the deprecated attribute.
15852 2018-06-15  Richard Biener  <rguenther@suse.de>
15854         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
15855         here, also noting vector size used.
15856         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
15857         size used in MSG_OPTIMIZED_LOCATIONS dump.
15858         (pass_slp_vectorize::execute): Adjust.
15860 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
15862         PR target/85968
15863         * config/arc/arc.c (arc_return_address_register): Fix
15864         if-condition.
15866 2018-06-15  Richard Biener  <rguenther@suse.de>
15868         PR middle-end/86159
15869         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
15870         leave useless conversion stripping to force_gimple_operand_gsi.
15871         (gimplify_build2): Likewise.
15872         (gimplify_build1): Likewise.
15874 2018-06-15  Richard Biener  <rguenther@suse.de>
15876         PR middle-end/86076
15877         * tree-cfg.c (move_stmt_op): unshare invariant addresses
15878         before adjusting their block.
15880 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
15882         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
15883         multilibs for *-*-rtems*.
15884         * config/riscv/t-rtems: New file.
15886 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
15888         PR middle-end/86122
15889         * match.pd ((A +- CST1) +- CST2): Punt if last resort
15890         unsigned_type_for returns NULL.
15892         PR target/85945
15893         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
15894         subregs of multi-word pseudos unless the float mode has word size.
15896 2018-06-14  Richard Biener  <rguenther@suse.de>
15898         PR middle-end/86139
15899         * tree-vect-generic.c (build_word_mode_vector_type): Remove
15900         duplicate and harmful type_hash_canon.
15901         * tree.c (type_hash_canon): Assert we didn't find ourselves.
15903 2018-06-14  Richard Biener  <rguenther@suse.de>
15905         PR ipa/86124
15906         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
15907         NULL cgraph_node.
15909 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
15911         * config/rtems.h (STDINT_LONG32): Define.
15913 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
15914             Prachi Godbole  <prachi.godbole@imgtec.com>
15916         * config/mips/mips-cpus.def: Define P6600.
15917         * config/mips/mips-tables.opt: Regenerate.
15918         * config/mips/mips.c (mips_ucbranch_type): New enum.
15919         (mips_rtx_cost_data): Add support for P6600.
15920         (mips_issue_rate): Likewise.
15921         (mips_multipass_dfa_lookahead): Likewise.
15922         (mips_avoid_hazard): Likewise.
15923         (mips_reorg_process_insns): Likewise.
15924         (mips_classify_branch_p6600): New function.
15925         * config/mips/mips.h (TUNE_P6600): New define.
15926         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
15927         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
15928         * config/mips/mips.md: Include p6600.md.
15929         (processor): Add p6600.
15930         * config/mips/p6600.md: New file.
15931         * doc/invoke.texi: Add p6600 to supported architectures.
15933 2018-06-13  Martin Sebor  <msebor@redhat.com>
15935         PR tree-optimization/86114
15936         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
15937         of integer types.
15938         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
15940 2018-06-13  Richard Biener  <rguenther@suse.de>
15942         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
15943         Properly set vector type of the intermediate stmt.
15944         * tree-vect-stmts.c (vectorizable_operation): The destination
15945         var always has vectype_out type.
15947 2018-06-13  Jeff Law  <law@redhat.com>
15949         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
15950         integer 0 for argument to print_rtl_with_bb.
15951         (rl78_reorg): Likewise.
15953 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
15955         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
15956         from rtx to rtx_insn *.
15957         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
15958         "label".
15959         (add_sched_insns_for_speculation): Likewise for local "target",
15960         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
15961         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
15962         from rtx_insn ** to rtx_code_label **.
15963         (reorg_emit_nops): Likewise.
15964         (c6x_reorg): Likewise for local "call_labels".
15965         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
15966         rtx to rtx_insn *.
15967         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
15968         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
15969         the loops over LABEL_REFS.
15970         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
15971         braf_label.
15972         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
15973         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
15974         (split_branches): Strengthen local "olabel" from rtx to
15975         rtx_insn *, adding a safe_as_a cast.
15976         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
15977         to "rtx_insn *".
15978         (add_insn_after): Likewise for first two params.
15979         (add_insn_before): Likewise.
15980         (remove_insn): Likewise for param.
15981         (emit_pattern_before_noloc): Likewise for second and third params.
15982         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
15983         (emit_call_insn_before_noloc): Likewise.
15984         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
15985         to "rtx_insn *".
15986         (emit_barrier_before): Likewise.
15987         (emit_label_before): Strengthen "label" param from "rtx" to
15988         "rtx_code_label *".  Strengthen "before" param from "rtx" to
15989         "rtx_insn *".
15990         (emit_insn_after_1): Strengthen "after" param from "rtx" to
15991         "rtx_insn *".
15992         (emit_pattern_after_noloc): Likewise.
15993         (emit_insn_after_noloc): Likewise.
15994         (emit_jump_insn_after_noloc): Likewise.
15995         (emit_call_insn_after_noloc): Likewise.
15996         (emit_debug_insn_after_noloc): Likewise.
15997         (emit_barrier_after): Likewise.
15998         (emit_label_after): Likewise for both params.
15999         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
16000         "loc" param from "int" to "location_t".
16001         (emit_insn_after_setloc): Likewise.
16002         (emit_jump_insn_after_setloc): Likewise.
16003         (emit_call_insn_after_setloc): Likewise.
16004         (emit_debug_insn_after_setloc): Likewise.
16005         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
16006         "loc" param from "int" to "location_t".
16007         (emit_pattern_before): Convert NULL_RTX to NULL.
16008         (emit_insn_before_setloc): Convert "loc" param from "int" to
16009         "location_t".
16010         (emit_jump_insn_before_setloc): Likewise.
16011         (emit_call_insn_before_setloc): Likewise.
16012         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
16013         rtx_insn *.  Convert "loc" param from "int" to "location_t".
16014         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
16015         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
16016         Convert 3rd param from "int" to "location_t".
16017         (emit_barrier_before, emit_barrier_after, next_real_insn):
16018         Strengthen param from rtx to rtx_insn *.
16019         (emit_label_before): Strengthen 1st param from "rtx" to
16020         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
16021         "rtx_insn *".
16022         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
16023         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
16024         Strengthen 2nd param from "rtx" to "rtx_insn *".
16025         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
16026         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
16027         Likewise. Convert 3rd param from "int" to "location_t".
16028         (emit_label_after): Strengthen 1st param from "rtx" to
16029         "rtx_code_label *".
16030         (next_real_insn, remove_insn): Strengthen param from "rtx" to
16031         "rtx_insn *".
16032         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
16033         from "rtx" to "rtx_insn *".
16035 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
16037         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
16038         bodies streamed in with -Q.
16039         * dumpfile.c (dump_files): Add lto-stream-out dump file.
16040         * dumpfile.h (tree_dump_index): Add lto_stream_out.
16041         * gimple-streamer-out.c: Include gimple-pretty-print.h
16042         (output_bb): Dump stmts streamed.
16043         * lto-section-out.c: Include print-tree.h
16044         (lto_begin_section): Dump sections created.
16045         (lto_output_decl_index): Dump decl encoded.
16046         * lto-streamer-out.c: Include print-tree.h
16047         (create_output_block): Dump output block created.
16048         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
16049         (output_function): Dump function output.
16050         (output_constructor): Dump constructor streamed.
16051         (write_global_stream): Output indexes encoded.
16052         (produce_asm_for_decls): Dump streams encoded.
16053         * lto-streamer.c (streamer_dump_file): New global var.
16054         * lto-streamer.h (streamer_dump_file): Declare.
16055         * passes.c (ipa_write_summaries): Initialize streamer dump.
16056         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
16057         in.
16059 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
16061         PR target/86048
16062         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
16063         offsets for register save directives.  Emit a second batch of save
16064         directives, if need be, when the function accesses prior frames.
16066 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16068         * config/arc/fpu.md (fmasf4): Force operand to register.
16069         (fnmasf4): Likewise.
16071 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16073         * config/arc/arc-protos.h (arc_pad_return): Remove.
16074         * config/arc/arc.c (machine_function): Remove force_short_suffix
16075         and size_reason.
16076         (arc_print_operand): Adjust printing of '&'.
16077         (arc_verify_short): Remove conditional printing of short suffix.
16078         (arc_final_prescan_insn): Remove reference to size_reason.
16079         (pad_return): New function.
16080         (arc_reorg): Call pad_return.
16081         (arc_pad_return): Remove.
16082         (arc_init_machine_status): Remove reference to force_short_suffix.
16083         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
16084         (attr length): When attribute iscompact is true force to 2
16085         regardless; in the case of maybe check if we want to force the
16086         instruction to have 4 bytes length.
16087         (nopv): Change it to generate 4 byte long nop as well.
16088         (blockage): New pattern.
16089         (simple_return): Remove call to arc_pad_return.
16090         (p_return_i): Likewise.
16092 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16094         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
16096 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16098         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
16099         ARC cores.
16101 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16103         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
16104         for ARC700 and ARCv2.
16106 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
16108         PR target/86076
16109         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
16110         operands[2] instead of operands[1].
16113 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
16115         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
16116         case, check whether the outer register overlaps an unallocatable
16117         register, not just whether it fits the required class.
16119 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
16121         * poly-int.h (can_div_trunc_p): Add new overload in which all values
16122         are poly_ints.
16123         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
16124         (memrefs_conflict_p): Likewise.
16125         (init_alias_analysis): Likewise.
16126         * cfgexpand.c (expand_debug_expr): Likewise.
16127         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
16128         * cse.c (fold_rtx): Likewise.
16129         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
16130         * expr.c (emit_block_move_hints): Likewise.
16131         (clear_storage_hints, push_block, emit_push_insn): Likewise.
16132         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
16133         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
16134         (emit_group_store): Likewise.
16135         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
16136         to read the PRE/POST_MODIFY increment.
16137         * calls.c (store_one_arg): Use strip_offset.
16138         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
16139         poly_int_rtx_p.
16140         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
16141         by a VEC_SELECT.
16142         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
16143         (simplify_binary_operation_1): Extend CONST_INT handling to
16144         poly_int_rtx_p.
16145         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
16146         than a HOST_WIDE_INT.
16147         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
16148         poly_int64.
16149         (adjust_mems, add_stores): Update accodingly.
16150         (vt_canonicalize_addr): Track polynomial offsets.
16151         (emit_note_insn_var_location): Likewise.
16152         (vt_add_function_parameter): Likewise.
16153         (vt_initialize): Likewise.
16155 2018-06-12  Jeff Law  <law@redhat.com>
16157         * config.gcc (alpha*-*-freebsd*): Remove.
16158         * config/alpha/freebsd.h: Remove.
16160 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
16162         PR other/69968
16163         * spellcheck-tree.c (levenshtein_distance): Rename to...
16164         (get_edit_distance): ...this, and update for underlying renaming.
16165         * spellcheck-tree.h (levenshtein_distance): Rename to...
16166         (get_edit_distance): ...this.
16167         * spellcheck.c (levenshtein_distance): Rename to...
16168         (get_edit_distance): ...this.  Convert from Levenshtein distance
16169         to Damerau-Levenshtein distance by supporting transpositions of
16170         adjacent characters.  Rename "v1" to "v_next" and "v0" to
16171         "v_one_ago".
16172         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
16173         (selftest::test_edit_distance_unit_test_oneway): ...this, and
16174         update for underlying renaming.
16175         (selftest::levenshtein_distance_unit_test): Rename to...
16176         (selftest::test_get_edit_distance_unit): ...this, and update for
16177         underlying renaming.
16178         (selftest::test_find_closest_string): Add example from PR 69968
16179         where transposition helps
16180         (selftest::test_metric_conditions): Update for renaming.
16181         (selftest::test_metric_conditions): Likewise.
16182         (selftest::spellcheck_c_tests): Likewise.
16183         * spellcheck.h (levenshtein_distance): Rename both overloads to...
16184         (get_edit_distance): ...this.
16185         (best_match::consider): Update for renaming.
16187 2018-06-12  Martin Sebor  <msebor@redhat.com>
16189         PR tree-optimization/85259
16190         * builtins.c (compute_objsize): Handle constant offsets.
16191         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
16192         true iff a warning has been issued.
16193         * gimple.h (gimple_nonartificial_location): New function.
16194         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
16195         gimple_nonartificial_location and handle -Wno-system-headers.
16196         (handle_builtin_stxncpy): Same.
16198 2018-06-12  Martin Sebor  <msebor@redhat.com>
16200         PR c/85931
16201         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
16203 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
16205         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
16206         BUILTIN_VEC_XST entries for pointer to double and long long.
16208 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
16210         PR target/85990
16211         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
16212         Update comments.
16213         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
16214         Likewise.
16216 2018-06-12  Martin Liska  <mliska@suse.cz>
16218         * doc/options.texi: Document IntegerRange.
16220 2018-06-12  Martin Liska  <mliska@suse.cz>
16222         * config/i386/i386.opt: Make MPX-related options as Deprecated.
16223         * opt-functions.awk: Handle Deprecated flag.
16224         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
16225         and report error.
16226         (read_cmdline_option): Report warning for a deprecated option.
16227         * opts.h (struct cl_option): Add new field cl_deprecated.
16228         (CL_ERR_DEPRECATED): New.
16230 2018-06-12  Martin Liska  <mliska@suse.cz>
16232         * doc/options.texi: Document Deprecated option flag.
16234 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16236         * config/arc/arc-arch.h (arc_extras): New enum.
16237         (arc_cpu_t):Add field extra.
16238         (arc_cpu_types): Consider the extras.
16239         * config/arc/arc-cpus.def: Add extras info.
16240         * config/arc/arc-opts.h (processor_type): Consider extra field.
16241         * config/arc/arc.c (arc_override_options): Handle extra field.
16243 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16245         * config/arc/arc-arch.h: Update ARC_OPTX macro.
16246         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
16247         field.
16248         * config/arc/arc.c (arc_init): Update pic warning.
16249         (irq_range): Update irq range parsing warnings.
16250         (arc_override_options): Update various warning messages.
16251         (arc_handle_aux_attribute): Likewise.
16253 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
16255         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
16257 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16259         * doc/sourcebuild.texi: Document usage of line number 0 in verify
16260         compiler messages directives.
16262 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
16264         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
16265         * config/mips/mips-tables.opt: Regenerate.
16266         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
16267         mips64r6.
16268         * doc/invoke.texi: Document -march=i6500.
16270 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
16272         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
16273         (i6400_gpmul): Add cpu_unit.
16274         (i6400_gpdiv): Likewise.
16275         (i6400_msa_add_d): Update reservations.
16276         (i6400_msa_int_add) Likewise.
16277         (i6400_msa_short_logic3) Likewise.
16278         (i6400_msa_short_logic2) Likewise.
16279         (i6400_msa_short_logic) Likewise.
16280         (i6400_msa_move) Likewise.
16281         (i6400_msa_cmp) Likewise.
16282         (i6400_msa_short_float2) Likewise.
16283         (i6400_msa_div_d) Likewise.
16284         (i6400_msa_long_logic1) Likewise.
16285         (i6400_msa_long_logic2) Likewise.
16286         (i6400_msa_mult) Likewise.
16287         (i6400_msa_long_float2) Likewise.
16288         (i6400_msa_long_float4) Likewise.
16289         (i6400_msa_long_float5) Likewise.
16290         (i6400_msa_long_float8) Likewise.
16291         (i6400_fpu_fadd): Include frint type.
16292         (i6400_fpu_store): New define_insn_reservation.
16293         (i6400_fpu_load): Likewise.
16294         (i6400_fpu_move): Likewise.
16295         (i6400_fpu_fcmp): Likewise.
16296         (i6400_fpu_fmadd): Likewise.
16297         (i6400_int_mult): Include imul3nc type and update reservation.
16298         (i6400_int_div): Include idiv3 type and update reservation.
16299         (i6400_int_load): Update to check type not move_type.
16300         (i6400_int_store): Likewise.
16301         (i6400_int_prefetch): Set zero latency.
16303 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
16305         * gcc.c: Document new %@{...} sequence.
16306         (LINK_COMMAND_SPEC): Use it for the -L switches.
16307         (cpp_unique_options): Use it for the -I switches.
16308         (at_file_argbuf): New global variable.
16309         (in_at_file): Likewise.
16310         (alloc_args): Create at_file_argbuf.
16311         (clear_args): Truncate at_file_argbuf.
16312         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
16313         (open_at_file): New function.
16314         (close_at_file): Likewise.
16315         (create_at_file): Delete.
16316         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
16317         <'o'>: Likewise.
16318         <'@'>: New case.
16319         (validate_switches_from_spec): Deal with %@{...} sequence.
16320         (validate_switches): Likewise.
16321         (driver::finalize): Call clear_args.
16323 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
16325         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
16327 2018-06-11  Martin Sebor  <msebor@redhat.com>
16329         * doc/invoke.texi (-Wall): List -Wc++17-compat.
16330         (Wno-class-memaccess): Add @opindex.
16331         (Wno-templates, Wno-multiple-inheritance): Same.
16332         (Wno-virtual-inheritance, Wno-namespaces): Same.
16333         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
16334         (Wno-format-overflow, Wno-format-truncation): Same.
16335         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
16336         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
16337         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
16338         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
16339         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
16340         (Wno-misspelled-isr): Same.
16342 2018-06-11  Martin Sebor  <msebor@redhat.com>
16344         * PR tree-optimization/86083
16345         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
16347 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
16349         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
16351 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
16353         PR target/85755
16354         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
16355         on the correct operand.
16356         (*movdi_internal64): Ditto.
16358 2018-06-11  Martin Liska  <mliska@suse.cz>
16360         PR tree-optimization/86089
16361         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
16363 2018-06-11  Julia Koval  <julia.koval@intel.com>
16365         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
16366         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
16367         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
16369 2018-06-11  Olivier Hainque  <hainque@adacore.com>
16371         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
16372         for Ada with strict dwarf2.
16374 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
16376         PR target/85755
16377         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
16378         addresses.
16380 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
16382         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
16384 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
16386         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
16387         TARGET_ELF.
16389 2018-06-08  Martin Liska  <mliska@suse.cz>
16391         * tree-cfg.h (debug_function): Fix argument type to match
16392         implementation.
16394 2018-06-08  Martin Liska  <mliska@suse.cz>
16396         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
16397         Remove usage of MPX-related (and removed) fields.
16398         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
16400 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
16402         * cfg.c (debug): Use TDF_NONE rather than 0.
16403         * cfghooks.c (debug): Likewise.
16404         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
16405         (struct dump_option_value_info): Convert to...
16406         (struct kv_pair): ...this template type.
16407         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
16408         rather than 0.
16409         (optinfo_verbosity_options): Likewise.
16410         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
16411         OPTGROUP_NONE.
16412         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
16413         than int for "optgroup_flags" param.
16414         (dump_generic_expr_loc): Use dump_flags_t rather than int for
16415         "dump_kind" param.
16416         (dump_dec): Likewise.
16417         (dump_finish): Use TDF_NONE rather than 0.
16418         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
16419         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
16420         than 0.  Update for change to option_ptr.
16421         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
16422         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
16423         0.  Update for changes to optinfo_verbosity_options and
16424         optgroup_options.
16425         (opt_info_switch_p): Convert optgroup_flags from int to
16426         optgroup_flags_t.
16427         (dump_basic_block): Use dump_flags_t rather than int
16428         for "dump_kind" param.
16429         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
16430         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
16431         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
16432         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
16433         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
16434         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
16435         TDF_NONE): Convert from macros to...
16436         (enum dump_flag): ...this new enum.
16437         (dump_flags_t): Update to use enum.
16438         (operator|, operator&, operator~, operator|=, operator&=):
16439         Implement for dump_flags_t.
16440         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
16441         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
16442         Convert from macros to...
16443         (enum optgroup_flag): ...this new enum.
16444         (optgroup_flags_t): New typedef.
16445         (operator|, operator|=): Implement for optgroup_flags_t.
16446         (struct dump_file_info): Convert field "alt_flags" to
16447         dump_flags_t.  Convert field "optgroup_flags" to
16448         optgroup_flags_t.
16449         (dump_basic_block): Use dump_flags_t rather than int for param.
16450         (dump_generic_expr_loc): Likewise.
16451         (dump_dec): Likewise.
16452         (dump_register): Convert param "optgroup_flags" to
16453         optgroup_flags_t.
16454         (opt_info_enable_passes): Likewise.
16455         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
16456         than 0.
16457         * gimple-pretty-print.c (debug): Likewise.
16458         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
16459         (merged_store_group::apply_stores): Likewise.
16460         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
16461         * gimple.c (verify_gimple_pp): Likewise.
16462         * graphite-poly.c (print_pbb_body): Likewise.
16463         * passes.c (pass_manager::register_one_dump_file): Convert
16464         local "optgroup_flags" to optgroup_flags_t.
16465         * print-tree.c (print_node): Use TDF_NONE rather than 0.
16466         (debug): Likewise.
16467         (debug_body): Likewise.
16468         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
16469         to optgroup_flags_t.
16470         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
16471         than 0.
16472         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
16473         (convert_mult_to_fma): Likewise.
16474         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
16475         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
16476         * tree-vect-data-refs.c (dump_lower_bound): Convert param
16477         "dump_kind" to dump_flags_t.
16479 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
16481         * config/rs6000/rs6000.c (min, max): Delete.
16483 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
16485         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
16486         -mabi=spe and -mabi=no-spe.
16488 2018-06-08  Martin Liska  <mliska@suse.cz>
16490         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
16491         where we expect an existing summary.
16493 2018-06-08  Martin Liska  <mliska@suse.cz>
16495         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
16496         * ipa-inline.h (estimate_edge_growth): Likewise.
16498 2018-06-08  Martin Liska  <mliska@suse.cz>
16500         * cgraph.c (function_version_hasher::hash): Use
16501         cgraph_node::get_uid ().
16502         (function_version_hasher::equal):
16503         * cgraph.h (cgraph_node::get_uid): New method.
16504         * ipa-inline.c (update_caller_keys): Use
16505         cgraph_node::get_uid ().
16506         (update_callee_keys): Likewise.
16507         * ipa-utils.c (searchc): Likewise.
16508         (ipa_reduced_postorder): Likewise.
16509         * lto-cgraph.c (input_node): Likewise.
16510         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
16511         * symbol-summary.h (symtab_insertion): Likewise.
16512         (symtab_removal): Likewise.
16513         (symtab_duplication): Likewise.
16514         * tree-pretty-print.c (dump_function_header): Likewise.
16515         * tree-sra.c (convert_callers_for_node): Likewise.
16517 2018-06-08  Martin Liska  <mliska@suse.cz>
16519         * cgraph.c (symbol_table::create_edge): Always assign a new
16520         unique number.
16521         (symbol_table::free_edge): Do not recycle numbers.
16522         * cgraph.h (cgraph_edge::get): New method.
16523         * symbol-summary.h (symtab_removal): Use it.
16524         (symtab_duplication): Likewise.
16525         (call_summary::hashable_uid): Remove.
16527 2018-06-08  Martin Liska  <mliska@suse.cz>
16529         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
16530         (initialize_growth_caches): Remove.
16531         (free_growth_caches): Likewise.
16532         (do_estimate_edge_time): Use edge_growth_cache.
16533         (do_estimate_edge_size): Likewise.
16534         (do_estimate_edge_hints): Likewise.
16535         * ipa-inline.c (reset_edge_caches): Likewise.
16536         (recursive_inlining): Likewise.
16537         (inline_small_functions): Likewise.
16538         * ipa-inline.h (initialize_growth_caches): Remove.
16539         (estimate_edge_size): Likewise.
16540         (estimate_edge_time): Likewise.
16541         (estimate_edge_hints): Likewise.
16542         (reset_edge_growth_cache): Likewise.
16543         * symbol-summary.h (call_summary::remove): New method.
16545 2018-06-08  Martin Liska  <mliska@suse.cz>
16547         * ipa-cp.c (class edge_clone_summary): New summary.
16548         (grow_edge_clone_vectors): Remove.
16549         (ipcp_edge_duplication_hook): Remove.
16550         (class edge_clone_summary_t): New call_summary class.
16551         (ipcp_edge_removal_hook): Remove.
16552         (edge_clone_summary_t::duplicate): New function.
16553         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
16554         (create_specialized_node): Likewise.
16555         (ipcp_driver): Initialize edge_clone_summaries and do not
16556         register hooks.
16558 2018-06-08  Martin Liska  <mliska@suse.cz>
16560         * symbol-summary.h (get): New function.
16561         (call_summary::m_initialize_when_cloning): New class member.
16563 2018-06-08  Martin Liska  <mliska@suse.cz>
16565         * cgraph.c (cgraph_node::remove): Do not recycle uid.
16566         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
16567         (symbol_table::allocate_cgraph_symbol): Do not set uid.
16568         * passes.c (uid_hash_t): Record removed_nodes by their uids.
16569         (remove_cgraph_node_from_order): Use the removed_nodes set.
16570         (do_per_function_toporder): Likwise.
16571         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
16572         instead of summary_uid.
16573         (symtab_removal): Likewise.
16574         (symtab_duplication): Likewise.
16576 2018-06-08  Martin Liska  <mliska@suse.cz>
16578         * ipa-cp.c (ipcp_store_bits_results): Use
16579         ipcp_transformation_sum.
16580         (ipcp_store_vr_results): Likewise.
16581         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
16582         to ...
16583         (ipcp_transformation_initialize): ... this.
16584         (ipa_set_node_agg_value_chain):
16585         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
16586         (write_ipcp_transformation_info): Likewise.
16587         (read_ipcp_transformation_info): Likewise.
16588         (ipcp_update_bits): Likewise.
16589         (ipcp_update_vr): Likewise.
16590         (ipcp_transform_function): Likewise.
16591         * ipa-prop.h: Rename ipcp_transformation_summary to
16592         ipcp_transformation.
16593         (class ipcp_transformation_t): New function summary.
16594         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
16595         (ipa_get_agg_replacements_for_node): Likewise.
16597 2018-06-08  Martin Liska  <mliska@suse.cz>
16599         * ipa-pure-const.c (struct funct_state_d): Do it class instead
16600         of struct.
16601         (class funct_state_summary_t): New function_summary class.
16602         (has_function_state): Remove.
16603         (get_function_state): Likewise.
16604         (set_function_state): Likewise.
16605         (add_new_function): Likewise.
16606         (funct_state_summary_t::insert): New function.
16607         (duplicate_node_data): Remove.
16608         (remove_node_data): Remove.
16609         (funct_state_summary_t::duplicate): New function.
16610         (register_hooks): Create new funct_state_summaries.
16611         (pure_const_generate_summary): Use it.
16612         (pure_const_write_summary): Likewise.
16613         (pure_const_read_summary): Likewise.
16614         (propagate_pure_const): Likewise.
16615         (propagate_nothrow): Likewise.
16616         (dump_malloc_lattice): Likewise.
16617         (propagate_malloc): Likewise.
16618         (execute): Do not register hooks, just remove summary
16619         instead.
16620         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
16621         constructor.
16623 2018-06-08  Martin Liska  <mliska@suse.cz>
16625         * ipa-reference.c (remove_node_data): Remove.
16626         (duplicate_node_data): Likewise.
16627         (class ipa_ref_var_info_summary_t): New class.
16628         (class ipa_ref_opt_summary_t): Likewise.
16629         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
16630         (get_reference_optimization_summary): Use
16631         ipa_ref_opt_sum_summaries.
16632         (set_reference_vars_info): Remove.
16633         (set_reference_optimization_summary): Likewise.
16634         (ipa_init): Create summaries.
16635         (init_function_info): Use function summary.
16636         (ipa_ref_opt_summary_t::duplicate): New function.
16637         (ipa_ref_opt_summary_t::remove): New function.
16638         (get_read_write_all_from_node): Fix GNU coding style.
16639         (propagate): Use function summary.
16640         (write_node_summary_p): Fix GNU coding style.
16641         (stream_out_bitmap): Likewise.
16642         (ipa_reference_read_optimization_summary): Use function summary.
16643         (ipa_reference_c_finalize): Do not release hooks.
16645 2018-06-08  Martin Liska  <mliska@suse.cz>
16647         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
16648         (analyze_function_body): Extract multiple calls of get_create.
16649         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
16650         * ipa-inline.c (recursive_inlining): Use ::get method.
16651         * ipa-inline.h (estimate_edge_growth): Likewise.
16653 2018-06-08  Martin Liska  <mliska@suse.cz>
16655         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
16656         HSA_INVALID.
16657         (hsa_function_summary::hsa_function_summary): Use the new enum
16658         value.
16659         (hsa_gpu_implementation_p): Use hsa_summaries::get.
16660         * hsa-gen.c (hsa_get_host_function): Likewise.
16661         (get_brig_function_name): Likewise.
16662         * ipa-hsa.c (process_hsa_functions): Likewise.
16663         (ipa_hsa_write_summary): Likewise.
16664         * symbol-summary.h (symtab_duplication): Use ::get function/
16665         (get): New function.
16667 2018-06-08  Martin Liska  <mliska@suse.cz>
16669         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
16670         of get.
16671         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
16672         (hsa_register_kernel): Likewise.
16673         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
16674         * hsa-gen.c (hsa_get_host_function): Likewise.
16675         (get_brig_function_name): Likewise.
16676         (generate_hsa): Likewise.
16677         (pass_gen_hsail::execute): Likewise.
16678         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
16679         (devirtualization_time_bonus): Likewise.
16680         (ipcp_propagate_stage): Likewise.
16681         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
16682         (edge_set_predicate): Likewise.
16683         (evaluate_conditions_for_known_args): Likewise.
16684         (evaluate_properties_for_edge): Likewise.
16685         (ipa_fn_summary::reset): Likewise.
16686         (ipa_fn_summary_t::duplicate): Likewise.
16687         (dump_ipa_call_summary): Likewise.
16688         (ipa_dump_fn_summary): Likewise.
16689         (analyze_function_body): Likewise.
16690         (compute_fn_summary): Likewise.
16691         (estimate_edge_devirt_benefit): Likewise.
16692         (estimate_edge_size_and_time): Likewise.
16693         (estimate_calls_size_and_time): Likewise.
16694         (estimate_node_size_and_time): Likewise.
16695         (inline_update_callee_summaries): Likewise.
16696         (remap_edge_change_prob): Likewise.
16697         (remap_edge_summaries): Likewise.
16698         (ipa_merge_fn_summary_after_inlining): Likewise.
16699         (ipa_update_overall_fn_summary): Likewise.
16700         (read_ipa_call_summary): Likewise.
16701         (inline_read_section): Likewise.
16702         (write_ipa_call_summary): Likewise.
16703         (ipa_fn_summary_write): Likewise.
16704         (ipa_free_fn_summary): Likewise.
16705         * ipa-hsa.c (process_hsa_functions): Likewise.
16706         (ipa_hsa_write_summary): Likewise.
16707         (ipa_hsa_read_section): Likewise.
16708         * ipa-icf.c (sem_function::merge): Likewise.
16709         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
16710         (do_estimate_edge_time): Likewise.
16711         (estimate_size_after_inlining): Likewise.
16712         (estimate_growth): Likewise.
16713         (growth_likely_positive): Likewise.
16714         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
16715         (inline_call): Likewise.
16716         * ipa-inline.c (caller_growth_limits): Likewise.
16717         (can_inline_edge_p): Likewise.
16718         (can_inline_edge_by_limits_p): Likewise.
16719         (compute_uninlined_call_time): Likewise.
16720         (compute_inlined_call_time): Likewise.
16721         (want_inline_small_function_p): Likewise.
16722         (edge_badness): Likewise.
16723         (update_caller_keys): Likewise.
16724         (update_callee_keys): Likewise.
16725         (recursive_inlining): Likewise.
16726         (inline_small_functions): Likewise.
16727         (inline_to_all_callers_1): Likewise.
16728         (dump_overall_stats): Likewise.
16729         (early_inline_small_functions): Likewise.
16730         (early_inliner): Likewise.
16731         * ipa-inline.h (estimate_edge_growth): Likewise.
16732         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
16733         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
16734         * ipa-prop.h (IPA_NODE_REF): Likewise.
16735         (IPA_EDGE_REF): Likewise.
16736         * ipa-pure-const.c (malloc_candidate_p): Likewise.
16737         (propagate_malloc): Likewise.
16738         * ipa-split.c (execute_split_functions): Likewise.
16739         * symbol-summary.h: Rename get to get_create.
16740         (get): Likewise.
16741         (get_create): Likewise.
16742         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
16744 2018-06-08  Martin Liska  <mliska@suse.cz>
16746         * symbol-summary.h (release): Move definition out of class
16747         declaration.
16748         (symtab_removal): Likewise.
16749         (symtab_duplication): Likewise.
16751 2018-06-08  Martin Liska  <mliska@suse.cz>
16753         * symbol-summary.h (function_summary): Move constructor
16754         implementation out of class declaration.
16755         (release): Likewise.
16756         (symtab_insertion): Likewise.
16757         (symtab_removal): Likewise.
16758         (symtab_duplication): Likewise.
16759         (get): Likewise.
16761 2018-06-08  Martin Liska  <mliska@suse.cz>
16763         * Makefile.in: Remove support for MPX (macros, related functions,
16764         fields in cgraph_node, ...).
16765         * builtin-types.def (BT_BND): Likewise.
16766         (BT_FN_BND_CONST_PTR): Likewise.
16767         (BT_FN_CONST_PTR_BND): Likewise.
16768         (BT_FN_VOID_PTR_BND): Likewise.
16769         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
16770         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
16771         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
16772         (expand_builtin_mempcpy_with_bounds): Likewise.
16773         (expand_builtin_memset_with_bounds): Likewise.
16774         (expand_builtin_memset_args): Likewise.
16775         (std_expand_builtin_va_start): Likewise.
16776         (expand_builtin): Likewise.
16777         (expand_builtin_with_bounds): Likewise.
16778         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
16779         (DEF_LIB_BUILTIN_CHKP): Likewise.
16780         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
16781         (DEF_CHKP_BUILTIN): Likewise.
16782         (BUILT_IN_MEMCPY): Likewise.
16783         (BUILT_IN_MEMMOVE): Likewise.
16784         (BUILT_IN_MEMPCPY): Likewise.
16785         (BUILT_IN_MEMSET): Likewise.
16786         (BUILT_IN_STPCPY): Likewise.
16787         (BUILT_IN_STRCAT): Likewise.
16788         (BUILT_IN_STRCHR): Likewise.
16789         (BUILT_IN_STRCPY): Likewise.
16790         (BUILT_IN_STRLEN): Likewise.
16791         (BUILT_IN_MEMCPY_CHK): Likewise.
16792         (BUILT_IN_MEMMOVE_CHK): Likewise.
16793         (BUILT_IN_MEMPCPY_CHK): Likewise.
16794         (BUILT_IN_MEMSET_CHK): Likewise.
16795         (BUILT_IN_STPCPY_CHK): Likewise.
16796         (BUILT_IN_STRCAT_CHK): Likewise.
16797         (BUILT_IN_STRCPY_CHK): Likewise.
16798         * calls.c (store_bounds): Likewise.
16799         (emit_call_1): Likewise.
16800         (special_function_p): Likewise.
16801         (maybe_warn_nonstring_arg): Likewise.
16802         (initialize_argument_information): Likewise.
16803         (finalize_must_preallocate): Likewise.
16804         (compute_argument_addresses): Likewise.
16805         (expand_call): Likewise.
16806         * cfgexpand.c (expand_call_stmt): Likewise.
16807         (expand_return): Likewise.
16808         (expand_gimple_stmt_1): Likewise.
16809         (pass_expand::execute): Likewise.
16810         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
16811         (cgraph_node::remove): Likewise.
16812         (cgraph_node::dump): Likewise.
16813         (cgraph_node::verify_node): Likewise.
16814         * cgraph.h (chkp_function_instrumented_p): Likewise.
16815         (symtab_node::get_alias_target): Likewise.
16816         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
16817         (cgraph_local_p): Likewise.
16818         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
16819         (cgraph_edge::rebuild_references): Likewise.
16820         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
16821         (walk_polymorphic_call_targets): Likewise.
16822         (cgraph_node::expand_thunk): Likewise.
16823         (symbol_table::output_weakrefs): Likewise.
16824         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
16825         (ix86_handle_option): Likewise.
16826         * config/i386/constraints.md: Likewise.
16827         * config/i386/i386-builtin-types.def (BND): Likewise.
16828         (VOID): Likewise.
16829         (PVOID): Likewise.
16830         (ULONG): Likewise.
16831         * config/i386/i386-builtin.def (BDESC_END): Likewise.
16832         (BDESC_FIRST): Likewise.
16833         (BDESC): Likewise.
16834         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
16835         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
16836         * config/i386/i386.c (enum reg_class): Likewise.
16837         (ix86_target_string): Likewise.
16838         (ix86_option_override_internal): Likewise.
16839         (ix86_conditional_register_usage): Likewise.
16840         (ix86_valid_target_attribute_inner_p): Likewise.
16841         (ix86_set_indirect_branch_type): Likewise.
16842         (ix86_set_current_function): Likewise.
16843         (ix86_function_arg_regno_p): Likewise.
16844         (init_cumulative_args): Likewise.
16845         (ix86_function_arg_advance): Likewise.
16846         (ix86_function_arg): Likewise.
16847         (ix86_pass_by_reference): Likewise.
16848         (ix86_function_value_regno_p): Likewise.
16849         (ix86_function_value_1): Likewise.
16850         (ix86_function_value_bounds): Likewise.
16851         (ix86_return_in_memory): Likewise.
16852         (ix86_setup_incoming_vararg_bounds): Likewise.
16853         (ix86_va_start): Likewise.
16854         (indirect_thunk_need_prefix): Likewise.
16855         (print_reg): Likewise.
16856         (ix86_print_operand): Likewise.
16857         (ix86_expand_call): Likewise.
16858         (ix86_output_function_return): Likewise.
16859         (reg_encoded_number): Likewise.
16860         (BDESC_VERIFYS): Likewise.
16861         (ix86_init_mpx_builtins): Likewise.
16862         (ix86_init_builtins): Likewise.
16863         (ix86_emit_cmove): Likewise.
16864         (ix86_emit_move_max): Likewise.
16865         (ix86_expand_builtin): Likewise.
16866         (ix86_builtin_mpx_function): Likewise.
16867         (ix86_get_arg_address_for_bt): Likewise.
16868         (ix86_load_bounds): Likewise.
16869         (ix86_store_bounds): Likewise.
16870         (ix86_load_returned_bounds): Likewise.
16871         (ix86_store_returned_bounds): Likewise.
16872         (ix86_class_likely_spilled_p): Likewise.
16873         (ix86_hard_regno_mode_ok): Likewise.
16874         (x86_order_regs_for_local_alloc): Likewise.
16875         (ix86_mitigate_rop): Likewise.
16876         (ix86_bnd_prefixed_insn_p): Likewise.
16877         (ix86_mpx_bound_mode): Likewise.
16878         (ix86_make_bounds_constant): Likewise.
16879         (ix86_initialize_bounds): Likewise.
16880         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
16881         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
16882         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
16883         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
16884         (TARGET_CHKP_BOUND_MODE): Likewise.
16885         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
16886         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
16887         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
16888         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
16889         * config/i386/i386.h (TARGET_MPX): Likewise.
16890         (TARGET_MPX_P): Likewise.
16891         (VALID_BND_REG_MODE): Likewise.
16892         (FIRST_BND_REG): Likewise.
16893         (LAST_BND_REG): Likewise.
16894         (enum reg_class): Likewise.
16895         (BND_REG_P): Likewise.
16896         (BND_REGNO_P): Likewise.
16897         (BNDmode): Likewise.
16898         (ADJUST_INSN_LENGTH): Likewise.
16899         * config/i386/i386.md: Likewise.
16900         * config/i386/i386.opt: Likewise.
16901         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
16902         (defined): Likewise.
16903         (LINK_MPX): Likewise.
16904         (MPX_SPEC): Likewise.
16905         (LIBMPX_SPEC): Likewise.
16906         (LIBMPXWRAPPERS_SPEC): Likewise.
16907         (CHKP_SPEC): Likewise.
16908         * config/i386/predicates.md: Likewise.
16909         * dbxout.c (dbxout_type): Likewise.
16910         * doc/extend.texi: Likewise.
16911         * doc/invoke.texi: Likewise.
16912         * doc/md.texi: Likewise.
16913         * doc/tm.texi: Likewise.
16914         * doc/tm.texi.in: Likewise.
16915         * dwarf2out.c (is_base_type): Likewise.
16916         (gen_formal_types_die): Likewise.
16917         (gen_subprogram_die): Likewise.
16918         (gen_type_die_with_usage): Likewise.
16919         (gen_decl_die): Likewise.
16920         (dwarf2out_late_global_decl): Likewise.
16921         * expr.c (expand_assignment): Likewise.
16922         (emit_storent_insn): Likewise.
16923         (store_expr_with_bounds): Likewise.
16924         (store_expr): Likewise.
16925         (expand_expr_real_1): Likewise.
16926         * expr.h (store_expr_with_bounds): Likewise.
16927         * function.c (use_register_for_decl): Likewise.
16928         (struct bounds_parm_data): Likewise.
16929         (assign_parms_augmented_arg_list): Likewise.
16930         (assign_parm_find_entry_rtl): Likewise.
16931         (assign_parm_is_stack_parm): Likewise.
16932         (assign_parm_load_bounds): Likewise.
16933         (assign_bounds): Likewise.
16934         (assign_parms): Likewise.
16935         (expand_function_start): Likewise.
16936         * gcc.c (CHKP_SPEC): Likewise.
16937         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16938         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
16939         (wrestrict_dom_walker::check_call): Likewise.
16940         * gimple.c (gimple_build_call_from_tree): Likewise.
16941         * gimple.h (enum gf_mask): Likewise.
16942         (gimple_call_with_bounds_p): Likewise.
16943         (gimple_call_set_with_bounds): Likewise.
16944         * gimplify.c (gimplify_init_constructor): Likewise.
16945         * ipa-cp.c (initialize_node_lattices): Likewise.
16946         (propagate_constants_across_call): Likewise.
16947         (find_more_scalar_values_for_callers_subset): Likewise.
16948         * ipa-hsa.c (process_hsa_functions): Likewise.
16949         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
16950         * ipa-icf.c (sem_function::merge): Likewise.
16951         * ipa-inline.c (early_inliner): Likewise.
16952         * ipa-pure-const.c (warn_function_noreturn): Likewise.
16953         (warn_function_cold): Likewise.
16954         (propagate_pure_const): Likewise.
16955         * ipa-ref.h (enum GTY): Likewise.
16956         * ipa-split.c (find_retbnd): Likewise.
16957         (consider_split): Likewise.
16958         (split_function): Likewise.
16959         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
16960         * ipa.c (walk_polymorphic_call_targets): Likewise.
16961         (symbol_table::remove_unreachable_nodes): Likewise.
16962         (process_references): Likewise.
16963         (cgraph_build_static_cdtor_1): Likewise.
16964         * lto-cgraph.c (lto_output_node): Likewise.
16965         (output_refs): Likewise.
16966         (compute_ltrans_boundary): Likewise.
16967         (input_overwrite_node): Likewise.
16968         (input_node): Likewise.
16969         (input_cgraph_1): Likewise.
16970         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
16971         * passes.c (pass_manager::execute_early_local_passes): Likewise.
16972         (class pass_chkp_instrumentation_passes): Likewise.
16973         (make_pass_chkp_instrumentation_passes): Likewise.
16974         * passes.def: Likewise.
16975         * rtl.h (struct GTY): Likewise.
16976         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
16977         * stor-layout.c (layout_type): Likewise.
16978         * symtab.c: Likewise.
16979         * target.def: Likewise.
16980         * targhooks.c (default_chkp_bound_type): Likewise.
16981         (default_chkp_bound_mode): Likewise.
16982         (default_builtin_chkp_function): Likewise.
16983         (default_chkp_function_value_bounds): Likewise.
16984         (default_chkp_make_bounds_constant): Likewise.
16985         (default_chkp_initialize_bounds): Likewise.
16986         * targhooks.h (default_chkp_bound_type): Likewise.
16987         (default_chkp_bound_mode): Likewise.
16988         (default_builtin_chkp_function): Likewise.
16989         (default_chkp_function_value_bounds): Likewise.
16990         (default_chkp_make_bounds_constant): Likewise.
16991         (default_chkp_initialize_bounds): Likewise.
16992         * toplev.c (compile_file): Likewise.
16993         (process_options): Likewise.
16994         * tree-core.h (DEF_BUILTIN): Likewise.
16995         (DEF_BUILTIN_CHKP): Likewise.
16996         * tree-inline.c (declare_return_variable): Likewise.
16997         (remap_gimple_stmt): Likewise.
16998         (copy_bb): Likewise.
16999         (initialize_inlined_parameters): Likewise.
17000         (expand_call_inline): Likewise.
17001         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
17002         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
17003         (make_pass_ipa_chkp_produce_thunks): Likewise.
17004         (make_pass_chkp): Likewise.
17005         (make_pass_chkp_opt): Likewise.
17006         (make_pass_chkp_instrumentation_passes): Likewise.
17007         * tree-pretty-print.c (dump_generic_node): Likewise.
17008         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
17009         * tree-ssa-dce.c (propagate_necessity): Likewise.
17010         (eliminate_unnecessary_stmts): Likewise.
17011         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
17012         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
17013         * tree-ssa-sccvn.h: Likewise.
17014         * tree-ssa-strlen.c (get_string_length): Likewise.
17015         (valid_builtin_call): Likewise.
17016         (adjust_last_stmt): Likewise.
17017         (handle_builtin_strchr): Likewise.
17018         (handle_builtin_strcpy): Likewise.
17019         (handle_builtin_stxncpy): Likewise.
17020         (handle_builtin_memcpy): Likewise.
17021         (handle_builtin_strcat): Likewise.
17022         (strlen_check_and_optimize_stmt): Likewise.
17023         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
17024         * tree-streamer-in.c: Likewise.
17025         * tree-streamer.c (record_common_node): Likewise.
17026         * tree.c (tree_code_size): Likewise.
17027         (wide_int_to_tree_1): Likewise.
17028         (type_contains_placeholder_1): Likewise.
17029         (build_common_tree_nodes): Likewise.
17030         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
17031         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
17032         (POINTER_BOUNDS_P): Likewise.
17033         (BOUNDED_TYPE_P): Likewise.
17034         (BOUNDED_P): Likewise.
17035         (CALL_WITH_BOUNDS_P): Likewise.
17036         (pointer_bounds_type_node): Likewise.
17037         * value-prof.c (gimple_ic): Likewise.
17038         * var-tracking.c (vt_add_function_parameters): Likewise.
17039         * varasm.c (make_decl_rtl): Likewise.
17040         (assemble_start_function): Likewise.
17041         (output_constant): Likewise.
17042         (maybe_assemble_visibility): Likewise.
17043         * varpool.c (ctor_for_folding): Likewise.
17044         * chkp-builtins.def: Remove.
17045         * ipa-chkp.c: Remove.
17046         * ipa-chkp.h: Remove.
17047         * rtl-chkp.c: Remove.
17048         * rtl-chkp.h: Remove.
17049         * tree-chkp-opt.c: Remove.
17050         * tree-chkp.c: Remove.
17051         * tree-chkp.h: Remove.
17053 2018-06-07  Carl Love  <cel@us.ibm.com>
17055         * config/rs6000/vsx.md (vextract_fp_from_shorth,
17056         vextract_fp_from_shortl): Add BE support.
17058 2018-06-07  Paul Koning  <ni1d@arrl.net>
17060         * compare-elim.c (try_merge_compare): Don't merge compare if
17061         address contains a side effect.
17062         (try_eliminate_compare): Likewise.
17064 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
17066         * config.gcc: Support "tremont".
17067         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
17068         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17069         PROCESSOR_TREMONT.
17070         * config/i386/i386.c (m_TREMONT): Define.
17071         (processor_target_table): Add "tremont".
17072         (PTA_TREMONT): Define.
17073         (ix86_lea_outperforms): Add TARGET_TREMONT.
17074         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
17075         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
17076         and M_INTEL_GOLDMONT_PLUS.
17077         (fold_builtin_cpu): Add "tremont".
17078         (ix86_add_stmt_cost): Add TARGET_TREMONT.
17079         (ix86_option_override_internal): Add "tremont".
17080         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
17081         (processor_type): Add PROCESSOR_TREMONT.
17082         * config/i386/x86-tune.def: Add m_TREMONT.
17083         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
17085 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
17087         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
17088         symbol defined for msp430i* devices to be lower case.
17090 2018-06-07  Richard Biener  <rguenther@suse.de>
17092         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
17093         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
17094         Properly wrap signed arithmetic if overflow wraps.
17096 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
17098         PR tree-optimization/69615
17099         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
17100         of a cast from a same precision integral SSA_NAME in a bb dominated
17101         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
17102         cast to utype if rhs2 has already a compatible type.
17104 2018-06-07  Richard Biener  <rguenther@suse.de>
17106         PR tree-optimization/85935
17107         * graphite-scop-detection.c (find_params_in_bb): Analyze
17108         condition operands with respect to the correct loop.  Assert
17109         the analysis doesn't fail.
17111 2018-06-04  Carl Love  <cel@us.ibm.com>
17113         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
17114         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
17115         as it is slightly cheaper.
17116         (first_match_or_eos_index_<mode>):
17117         Calculate index using natural element order.
17118         (first_match_index_<mode>):
17119         Calculate index using natural element order.
17120         (first_match_or_eos_index_<mode>):
17121         Calculate index using natural order.
17122         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
17123         for BE and LE modes.
17124         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
17125         P9V_BUILTIN_VCLZLSBB_V16QI.
17126         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
17127         specific.
17129 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17131         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
17132         indentation and line wrap for many prototypes.  Add missing
17133         @smallexample directives around block of prototypes for vec_xl and
17134         vec_xst.
17136 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
17138         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
17139         track if we pass or return IEEE 128-bit floating point.
17140         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
17141         C++ mangling that is compatible with GCC 8.1.
17142         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
17143         (init_cumulative_args): Note if we pass or return IEEE 128-bit
17144         floating point types.
17145         (rs6000_function_arg_advance_1): Likewise.
17146         (rs6000_mangle_type): Optionally generate mangled names that match
17147         what GCC 8.1 generated for IEEE 128-bit floating point types.
17148         (rs6000_globalize_decl_name): If we have an external function that
17149         passes or returns IEEE 128-bit types, generate a weak reference
17150         from the mangled name used in GCC 8.1 to the current mangled
17151         name.
17152         (rs6000_init_builtins): Make __ibm128 use the long double type if
17153         long double is IBM extended double.  Make __float128 use the long
17154         double type if long double is IEEE 128-bit.
17156         PR target/85657
17157         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
17158         macro for __ibm128 built-in functions.
17159         (PACK_IF): Add __ibm128 pack/unpack functions.
17160         (UNPACK_IF): Likewise.
17161         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
17162         enable long double built-in functions if long double is IEEE
17163         128-bit floating point.
17164         (rs6000_invalid_builtin): Update long double built-in function
17165         error message.
17166         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
17167         functions, adjust the built-in function to use the long double
17168         built-in function if __ibm128 and long double are the same type.
17169         * doc/extend.texi (PowerPC builtins): Update documention for
17170         __builtin_{,un}pack_longdouble.  Add documentation for
17171         __builtin_{,un}pack_ibm128.
17173 2018-06-06  Jim Wilson  <jimw@sifive.com>
17175         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
17176         (struct machine_function): New field interrupt_mode.
17177         (riscv_handle_type_attribute): New function.  Add forward declaration.
17178         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
17179         (riscv_expand_epilogue): Check interrupt_mode field.
17180         (riscv_set_current_function): Check interrupt attribute args and
17181         set interrupt_mode field.
17182         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
17183         (riscv_sret, riscv_uret): New.
17184         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
17185         new arguments to interrupt attribute.
17187 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
17189         PR target/63177
17190         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
17191         Don't handle -mcpu=power8 if -mpower9-vector is also used.
17193 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17195         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
17196         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
17197         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
17198         several redundant entries.
17200 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
17202         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
17203         type from "rtx" to "rtx_insn *".
17204         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
17205         for local "call_insn", removing cast.
17206         (ix86_expand_call): Likewise, introducing a "call_insn" local.
17208 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
17210         PR tree-optimization/86066
17211         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
17212         for BIT_INSERT_EXPR stores.
17214 2018-06-06  Richard Biener  <rguenther@suse.de>
17216         PR tree-optimization/86062
17217         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
17218         component refs ontop
17219         of to be offsetted base.
17221 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
17223         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
17224         to be static and remove check on interrupt attribute name.
17226 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17228         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
17229         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
17231 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
17233         PR target/79924
17234         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
17235         second argument.
17236         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
17237         Remove second argument, change how error is called.
17238         (aarch64_layout_arg): Remove second argument from
17239         aarch64_err_no_fpadvsimd call.
17240         (aarch64_init_cumulative_args): Ditto.
17241         (aarch64_gimplify_va_arg_expr): Ditto.
17242         * config/aarch64/aarch64.md (mov<mode>): Ditto.
17244 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
17246         * config/i386/i386.md (simple_return_indirect_internal): New expander.
17247         (*simple_return_indirect_internal<mode>): Rename from
17248         simple_return_indirect_internal.  Use W mode iterator.
17249         (rstorssp): New expander.
17250         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
17251         (clrssbsy): New expander.
17252         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
17254 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
17256         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
17257         __typeof__.
17258         (cmse_check_pointed_object): Likewise.
17260 2018-06-05  Martin Liska  <mliska@suse.cz>
17262         PR gcov-profile/47618
17263         * doc/invoke.texi: Document how -fprofile-dir format
17264         is extended.
17266 2018-06-05  Richard Biener  <rguenther@suse.de>
17268         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
17269         removal pretend DOM info isn't available so we do not update
17270         it and only remove edges, not dominated blocks.  Actually free
17271         DOM info in case we removed something.  Remove unreachable blocks.
17272         (mfb_keep_latches): Work with either DOM info or marked backedges.
17273         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
17274         first.  Mark backedges if DOM info isn't available.
17275         (Re-)compute DOM info after cleanup_control_flow_pre.
17277 2018-06-05  Richard Biener  <rguenther@suse.de>
17279         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
17280         (locus_discrim_hasher::hash): Adjust.
17281         (locus_discrim_hasher::equal): Likewise.
17282         (next_discriminator_for_locus): Work on line directly.
17283         (same_line_p): Pass in expanded locus1 as well.
17284         (assign_discriminators): Avoid redundant location expansions.
17286 2018-06-05  Richard Biener  <rguenther@suse.de>
17288         PR tree-optimization/86046
17289         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
17290         if required after clearing TREE_ADDRESSABLE.
17292 2018-06-05  Richard Biener  <rguenther@suse.de>
17294         PR tree-optimization/86047
17295         * tree-ssa-loop.c (for_each_index): Glob handling of all
17296         decls and constants and really handle all of them.
17298 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17300         PR target/81497
17301         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
17302         qualifier_void_pointer and qualifier_const_void_pointer.
17303         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
17304         (arm_init_builtins): Handle the above.
17305         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
17306         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
17307         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
17308         void intrinsics.
17310 2018-06-05  Martin Liska  <mliska@suse.cz>
17312         * auto-profile.c (read_autofdo_file): Do not use
17313         gcov_ctr_summary struct.
17314         (afdo_callsite_hot_enough_for_early_inline): Likewise.
17315         * coverage.c (struct counts_entry): Likewise.
17316         (read_counts_file): Read just single summary entry.
17317         (get_coverage_counts): Use gcov_summary struct.
17318         * coverage.h (get_coverage_counts): Likewise.
17319         * gcov-dump.c (dump_working_sets): Likewise.
17320         (tag_summary): Dump just single summary.
17321         * gcov-io.c (gcov_write_summary): Write just histogram
17322         summary.
17323         (gcov_read_summary): Read just single summary.
17324         (compute_working_sets): Use gcov_summary struct.
17325         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
17326         of GCOV_COUNTERS_SUMMABLE.
17327         (GCOV_COUNTERS_SUMMABLE): Remove.
17328         (GCOV_FIRST_VALUE_COUNTER): Replace with
17329         GCOV_COUNTER_V_INTERVAL.
17330         (struct gcov_ctr_summary): Remove.
17331         (struct gcov_summary): Directly use fields of former
17332         gcov_ctr_summary.
17333         (compute_working_sets): Use gcov_summary struct.
17334         * gcov.c (read_count_file): Do not use ctrs fields.
17335         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
17336         struct.
17337         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
17338         struct.
17339         * profile.c: Likewise.
17340         * profile.h: Likewise.
17342 2018-06-05  Martin Liska  <mliska@suse.cz>
17344         PR gcov-profile/84846
17345         * gcov.c (output_lines): Print working directory only
17346         in intermediate format.
17348 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
17350         * config/s390/s390-builtin-types.def: Add void function type.
17351         * config/s390/s390-builtins.def: Use the function type for the
17352         tbeginc builtin.
17354 2018-06-04  Jim Wilson  <jimw@sifive.com>
17356         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
17357         to int.
17358         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
17359         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
17360         handle EH_RETURN_DATA_REGNO registers properly.
17361         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
17362         (riscv_expand_epilogue): Update comment.  Change argument name and
17363         type.  Update code to use new name and type.  Pass new args to
17364         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
17365         EXCEPTION_RETURN.
17366         * config/riscv/riscv.md (NORMAL_RETURN): New.
17367         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
17368         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
17369         (eh_return): Call gen_eh_return_internal and emit barrier.
17370         (eh_return_internal): Call riscv_expand_epilogue.
17372 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
17374         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
17375         bit_insertion field and declare can_be_merged_into method.
17376         (merged_store_group::can_be_merged_into): New method.
17377         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
17378         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
17379         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
17381 2018-06-04  Richard Biener  <rguenther@suse.de>
17383         PR tree-optimization/85955
17384         * builtins.c (fold_builtin_sincos): Convert pointers to
17385         destination to appropriate type before dereferencing.
17387 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
17389         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
17391 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
17393         * expr.c (expand_expr_real_1): Force the operand into memory if
17394         its TYPE_MODE is BLKmode and if there is no integer mode for
17395         the number of bits being extracted.
17397 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
17399         PR target/85832
17400         PR target/86036
17401         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
17402         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
17404 2018-06-04  Richard Biener  <rguenther@suse.de>
17406         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
17407         (cleanup_tree_cfg_noloop): ... single caller.  Do
17408         start_recording_case_labels later.
17410 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
17412         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
17413         to _IMMINTRIN_H_INCLUDED.
17414         * config/i386/pconfigintrin.h: Ditto.
17415         * config/i386/waitpkgintrin.h: Ditto.
17416         * config/i386/immintrin.h: Add includes for sgxintrin.h,
17417         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
17418         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
17419         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
17420         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
17421         waitpkgintrin.h and cldemoteintrin.h.
17423 2018-06-04  Richard Biener  <rguenther@suse.de>
17425         PR tree-optimization/86038
17426         * tracer.c (find_best_successor): Check probability for
17427         being initialized, bail out if not.
17429 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
17431         PR target/86003
17432         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
17433         of bits to ignore when comparing architectures.
17435 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
17437         PR tree-optimization/69615
17438         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
17439         maximum or minimum of the type, try to merge it also as if
17440         range1 is + [-, x - 1] or + [x + 1, -].
17442         PR c++/86025
17443         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
17445 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
17447         PR tree-optimization/86034
17448         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
17449         the unsigned bitfield type in a bit insertion sequence if it does not
17450         have a larger precision than the bitfield size.
17451         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
17453 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
17455         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
17457 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
17459         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
17460         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
17461         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
17462         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
17464 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
17466         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
17467         Disable -fdelete-null-pointer-checks for ELF toolchain.
17469 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
17470             Kito Cheng  <kito.cheng@gmail.com>
17472         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
17473         (nds32le-*-*, nds32be-*-*): Integrate checking process.
17474         (nds32*-*-*): Add glibc and uclibc conditions.
17475         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
17476         (TARGET_EXCEPT_UNWIND_INFO): Define.
17477         * config/nds32/elf.h: New file.
17478         * config/nds32/linux.h: New file.
17479         * config/nds32/nds32-elf.opt: New file.
17480         * config/nds32/nds32-linux.opt: New file.
17481         * config/nds32/nds32-fp-as-gp.c
17482         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
17483         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
17484         TARGET_LINUX_ABI.
17485         (nds32_asm_file_end): Ditto.
17486         (nds32_print_operand): Ditto.
17487         (nds32_insert_attributes): Ditto.
17488         (nds32_init_libfuncs): New function.
17489         (TARGET_HAVE_TLS): Define.
17490         (TARGET_INIT_LIBFUNCS): Define.
17491         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
17492         spec content.
17493         (TARGET_ELF): Apply different mcmodel setting.
17494         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
17495         been migrated into elf.h and linux.h files.
17496         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
17497         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
17498         (mcmodel): The content has been migrated into nds32-elf.opt and
17499         nds32-linux.opt files.
17500         * config/nds32/t-elf: New file.
17501         * config/nds32/t-linux: New file.
17503 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
17504             Shiva Chen  <shiva0217@gmail.com>
17506         * config/nds32/constants.md (unspec_volatile_element): Add
17507         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
17508         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
17509         optimization.
17510         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
17511         (make_pass_nds32_fp_as_gp): Declare.
17512         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
17513         optmization pass.
17514         (nds32_asm_function_end_prologue): Remove unused asm output.
17515         (nds32_asm_function_begin_epilogue): Remove unused asm output.
17516         (nds32_asm_file_start): Output necessary fp_as_gp information.
17517         (nds32_option_override): Adjust register usage.
17518         (nds32_expand_prologue): Consider fp_as_gp situation.
17519         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
17520         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
17521         (epilogue): Ditto.
17522         (return): Ditto.
17523         (simple_return): Ditto.
17524         (omit_fp_begin): Output special directive for fp_as_gp.
17525         (omit_fp_end): Output special directive for fp_as_gp.
17526         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
17527         mforbid-fp-as-gp): New options.
17529 2018-06-01  Mark Wielaard  <mark@klomp.org>
17531         * dwarf2out.c (dwarf2out_finish): Remove generation of
17532         DW_AT_loclists_base.
17534 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
17536         * gimple-ssa-store-merging.c: Include gimple-fold.h.
17537         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
17538         (struct merged_store_group): Add bit_insertion field.
17539         (dump_char_array): Use standard hexadecimal format.
17540         (merged_store_group::merged_store_group): Set bit_insertion to false.
17541         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
17542         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
17543         also print the mask in the dump file.
17544         (pass_store_merging::gate): Minor tweak.
17545         (imm_store_chain_info::coalesce_immediate): Fix wrong association
17546         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
17547         stores with INTEGER_CST stores.
17548         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
17549         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
17550         and use it throughout.  Generate bit insertion sequences if need be.
17551         (pass_store_merging::process_store): Remove redundant condition.
17552         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
17554 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
17556         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
17557         the 128-bit floating point types.  Fix function comment.
17559 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17561         * config/aarch64/aarch64-simd.md
17562         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
17563         mnemonics.
17564         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
17565         mnemonics.
17567 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
17569         PR tree-optimization/85989
17570         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
17571         variable.
17572         (backprop::intersect_uses): Check it when deciding whether this
17573         is a backedge reference.
17574         (backprop::process_block): Add each phi to m_visited_phis
17575         after visiting it, then clear it at the end.
17577 2018-06-01  Richard Biener  <rguenther@suse.de>
17579         * tree-vectorizer.h (vect_dr_stmt): New function.
17580         (vect_get_load_cost): Adjust.
17581         (vect_get_store_cost): Likewise.
17582         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
17583         Use vect_dr_stmt instead of DR_SMTT.
17584         (vect_record_base_alignments): Likewise.
17585         (vect_calculate_target_alignment): Likewise.
17586         (vect_compute_data_ref_alignment): Likewise and make static.
17587         (vect_update_misalignment_for_peel): Likewise.
17588         (vect_verify_datarefs_alignment): Likewise.
17589         (vector_alignment_reachable_p): Likewise.
17590         (vect_get_data_access_cost): Likewise.  Pass down
17591         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
17592         (vect_get_peeling_costs_all_drs): Likewise.
17593         (vect_peeling_hash_get_lowest_cost): Likewise.
17594         (vect_enhance_data_refs_alignment): Likewise.
17595         (vect_find_same_alignment_drs): Likewise.
17596         (vect_analyze_data_refs_alignment): Likewise.
17597         (vect_analyze_group_access_1): Likewise.
17598         (vect_analyze_group_access): Likewise.
17599         (vect_analyze_data_ref_access): Likewise.
17600         (vect_analyze_data_ref_accesses): Likewise.
17601         (vect_vfa_segment_size): Likewise.
17602         (vect_small_gap_p): Likewise.
17603         (vectorizable_with_step_bound_p): Likewise.
17604         (vect_prune_runtime_alias_test_list): Likewise.
17605         (vect_analyze_data_refs): Likewise.
17606         (vect_supportable_dr_alignment): Likewise.
17607         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
17608         (vect_gen_prolog_loop_niters): Likewise.
17609         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
17610         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
17611         modify DR_STMT.
17612         (vect_recog_mask_conversion_pattern): Likewise.
17613         (vect_try_gather_scatter_pattern): Likewise.
17614         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
17615         to vect_get_store_cost.
17616         (vect_get_store_cost): Get stmt_info instead of DR.
17617         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
17618         (vect_get_load_cost): Get stmt_info instead of DR.
17620 2018-06-01  Richard Biener  <rguenther@suse.de>
17622         PR middle-end/86017
17623         * gimple-fold.c (var_decl_component_p): Also allow offsetted
17624         vars wrapped in MEM_REFs.
17626 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
17628         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
17629         Fix subreg tests so that we only return a choice between
17630         GENERAL_REGS and FP_REGS if the original classes included both.
17632 2018-06-01  Richard Biener  <rguenther@suse.de>
17634         PR ipa/85960
17635         * tree-ssa-structalias.c (get_function_part_constraint):
17636         Handle NULL fi->decl.
17637         (find_func_aliases_for_call): Properly handle indirect
17638         fi from direct call.
17639         (find_func_clobbers): Likewise.
17640         (ipa_pta_execute): Likewise.
17641         (create_variable_info_for): For functions that are ifunc_resolver
17642         resolve to a varinfo that contains the result of the resolver call.
17643         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
17644         aliases.
17646 2018-05-31  Michael Collison  <michael.collison@arm.com>
17648         * config/aarch64/aarch64.md:
17649         (*fix_to_zero_extenddfdi2): New pattern.
17650         * gcc.target/aarch64/fix_extend1.c: New testcase.
17652 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
17654         PR middle-end/78809
17655         PR middle-end/83026
17656         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
17657         and BUILT_IN_STRNCMP_EQ.
17658         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
17659         BUILT_IN_STRNCMP_EQ.
17660         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
17661         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
17662         (gimple_fold_builtin): Likewise.
17663         * tree-ssa-strlen.c (compute_string_length): New function.
17664         (determine_min_obsize): New function.
17665         (handle_builtin_string_cmp): New function to handle calls to
17666         string compare functions.
17667         (strlen_optimize_stmt): Add handling to builtin string compare
17668         calls.
17669         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
17670         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
17671         * tree.c (build_common_builtin_nodes): Add new defines of
17672         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
17674 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
17676         PR target/85984
17677         * bb-reorder.c (pass_partition_blocks::gate): Return false for
17678         functions with naked attribute.
17680 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
17682         * config/i386/sse.md (avx_vec_concat<mode>):
17683         Substitute concat_tg_mode mode attribute with xtg_mode.
17684         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
17685         (concat_tg_mode): Remove mode attribute.
17687 2018-05-31  Martin Sebor  <msebor@redhat.com>
17689         PR c/82063
17690         * calls.c (alloc_max_size): Correct a logic error/typo.
17691         Treat excessive arguments as infinite.  Warn for invalid arguments.
17692         * doc/invoke.texi (-Walloc-size-larger-than): Update.
17694 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
17696         PR target/85829
17697         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
17698         and movx for Haswell.
17700 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
17701             Cesar Philippidis  <cesar@codesourcery.com>
17703         PR middle-end/85879
17704         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
17705         when emitting error on private/firstprivate reductions.
17706         * omp-low.c (lower_omp_target): Avoid reference-type processing
17707         on pointers for firstprivate clause.
17709 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
17711         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
17712         (st1x2): Likewise.
17713         (st1x3): Likewise.
17714         * config/aarch64/aarch64-simd.md
17715         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
17716         (aarch64_ld1_x3_<mode>): Likewise
17717         (aarch64_st1x2<VALLDIF:mode>): Likewise
17718         (aarch64_st1_x2_<mode>): Likewise
17719         (aarch64_st1x3<VALLDIF:mode>): Likewise
17720         (aarch64_st1_x3_<mode>): Likewise
17721         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
17722         (vld1_s8_x3): Likewise.
17723         (vld1_u16_x3): Likewise.
17724         (vld1_s16_x3): Likewise.
17725         (vld1_u32_x3): Likewise.
17726         (vld1_s32_x3): Likewise.
17727         (vld1_u64_x3): Likewise.
17728         (vld1_s64_x3): Likewise.
17729         (vld1_f16_x3): Likewise.
17730         (vld1_f32_x3): Likewise.
17731         (vld1_f64_x3): Likewise.
17732         (vld1_p8_x3): Likewise.
17733         (vld1_p16_x3): Likewise.
17734         (vld1_p64_x3): Likewise.
17735         (vld1q_u8_x3): Likewise.
17736         (vld1q_s8_x3): Likewise.
17737         (vld1q_u16_x3): Likewise.
17738         (vld1q_s16_x3): Likewise.
17739         (vld1q_u32_x3): Likewise.
17740         (vld1q_s32_x3): Likewise.
17741         (vld1q_u64_x3): Likewise.
17742         (vld1q_s64_x3): Likewise.
17743         (vld1q_f16_x3): Likewise.
17744         (vld1q_f32_x3): Likewise.
17745         (vld1q_f64_x3): Likewise.
17746         (vld1q_p8_x3): Likewise.
17747         (vld1q_p16_x3): Likewise.
17748         (vld1q_p64_x3): Likewise.
17749         (vst1_s64_x2): Likewise.
17750         (vst1_u64_x2): Likewise.
17751         (vst1_f64_x2): Likewise.
17752         (vst1_s8_x2): Likewise.
17753         (vst1_p8_x2): Likewise.
17754         (vst1_s16_x2): Likewise.
17755         (vst1_p16_x2): Likewise.
17756         (vst1_s32_x2): Likewise.
17757         (vst1_u8_x2): Likewise.
17758         (vst1_u16_x2): Likewise.
17759         (vst1_u32_x2): Likewise.
17760         (vst1_f16_x2): Likewise.
17761         (vst1_f32_x2): Likewise.
17762         (vst1_p64_x2): Likewise.
17763         (vst1q_s8_x2): Likewise.
17764         (vst1q_p8_x2): Likewise.
17765         (vst1q_s16_x2): Likewise.
17766         (vst1q_p16_x2): Likewise.
17767         (vst1q_s32_x2): Likewise.
17768         (vst1q_s64_x2): Likewise.
17769         (vst1q_u8_x2): Likewise.
17770         (vst1q_u16_x2): Likewise.
17771         (vst1q_u32_x2): Likewise.
17772         (vst1q_u64_x2): Likewise.
17773         (vst1q_f16_x2): Likewise.
17774         (vst1q_f32_x2): Likewise.
17775         (vst1q_f64_x2): Likewise.
17776         (vst1q_p64_x2): Likewise.
17777         (vst1_s64_x3): Likewise.
17778         (vst1_u64_x3): Likewise.
17779         (vst1_f64_x3): Likewise.
17780         (vst1_s8_x3): Likewise.
17781         (vst1_p8_x3): Likewise.
17782         (vst1_s16_x3): Likewise.
17783         (vst1_p16_x3): Likewise.
17784         (vst1_s32_x3): Likewise.
17785         (vst1_u8_x3): Likewise.
17786         (vst1_u16_x3): Likewise.
17787         (vst1_u32_x3): Likewise.
17788         (vst1_f16_x3): Likewise.
17789         (vst1_f32_x3): Likewise.
17790         (vst1_p64_x3): Likewise.
17791         (vst1q_s8_x3): Likewise.
17792         (vst1q_p8_x3): Likewise.
17793         (vst1q_s16_x3): Likewise.
17794         (vst1q_p16_x3): Likewise.
17795         (vst1q_s32_x3): Likewise.
17796         (vst1q_s64_x3): Likewise.
17797         (vst1q_u8_x3): Likewise.
17798         (vst1q_u16_x3): Likewise.
17799         (vst1q_u32_x3): Likewise.
17800         (vst1q_u64_x3): Likewise.
17801         (vst1q_f16_x3): Likewise.
17802         (vst1q_f32_x3): Likewise.
17803         (vst1q_f64_x3): Likewise.
17804         (vst1q_p64_x3): Likewise.
17806 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
17808         * config/msp430/msp430.c (msp430_output_labelref): Prepend
17809         user_label_prefix to name.
17811         * tree-core.h: Update comment about the format of NAME string
17812         passed to handler in attribute_spec.
17814         * config/msp430/msp430.md: Remove erroneous subreg expression from
17815         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
17816         zero_extend{q,h}isi2.
17818 2018-05-30  Borislav Petkov  <bp@suse.de>
17820         * doc/extend.texi: Document some architecture specific
17821         constraints and sort entries.
17823 2018-05-30  Martin Sebor  <msebor@redhat.com>
17825         PR middle-end/85369
17826         * builtins.c (expand_builtin_stpcpy_1): New function.
17827         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
17828         only if the former succeeds.
17830 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
17832         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
17833         in saphira.
17835 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17837         * doc/invoke.texi (-flinker-output): Document
17839 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17841         * passes.c (ipa_write_summaries): Only modify statements if body
17842         is in memory.
17843         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
17844         incrementally linking.
17845         (ipa_passes): Likewise.
17846         * lto-cgraph.c (lto_output_node): When incrementally linking do not
17847         pass down resolution info.
17848         * common.opt (flag_incremental_link): Update info.
17849         * gcc.c (plugin specs): Turn flinker-output=* to
17850         -plugin-opt=-linker-output-known
17851         * toplev.c (compile_file): Also cut compilation when doing incremental
17852         link.
17853         * flag-types. (enum lto_partition_model): Add
17854         LTO_LINKER_OUTPUT_NOLTOREL.
17855         (invoke.texi): Add -flinker-output docs.
17856         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
17857         link same way as WPA; do not stream in dead initializers.
17859         * dwarf2out.c (dwarf2out_die_ref_for_decl,
17860         darf2out_register_external_decl): Support incremental link.
17862 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17864         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
17866 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17868         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
17869         it down to simple_object_copy_lto_debug_sections.
17870         (run_gcc): Determine incremental LTO link time and configure
17871         lto1 into non-wpa mode, disable renaming of debug sections.
17873 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17875         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
17876         descriptions of various incorrectly documented functions.
17878 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
17880         Revert:
17881         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
17882         address check not strict.
17884 2018-05-30  Richard Biener  <rguenther@suse.de>
17886         PR tree-optimization/85964
17887         * tracer.c (better_p): Drop initialized count check, we only
17888         call the function with initialized counts now.
17889         (find_best_successor): Do find a best edge if one
17890         has uninitialized count.
17891         (find_best_predecessor): Likewise.  Do BB frequency check only
17892         if count is initialized.
17894 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
17896         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
17897         (aarch64_ldrstr_offset_compare): New.
17898         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
17899         load/store orderings.
17900         (aarch64_gen_adjusted_ldpstp): Likewise.
17902 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
17904         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
17905         Check for subset of GENERAL_REGS and FP_REGS.
17906         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
17907         r=w alternative.
17909 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
17911         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
17912         and wi::to_poly_offset.  Add the current offset and then check
17913         whether the sum fits, rather than using an unchecked addition of
17914         a checked term.  Check for a shwi rather than a uhwi.
17915         * expr.c (get_bit_range): Use tree_to_poly_uint64.
17916         (store_constructor): Use poly_int_tree_p.
17917         (expand_expr_real_1): Likewise.
17918         * function.c (assign_temp): Likewise.
17919         * fold-const.c (const_binop): Use poly_int_tree_p and
17920         wi::to_poly_offset.
17921         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
17922         division.
17923         * ipa-icf-gimple.c (func_checker::compare_operand): Use
17924         to_poly_offset for MEM offsets.
17925         * ipa-icf.c (sem_variable::equals): Likewise.
17926         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
17927         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
17928         wi::to_poly_offset for BIT_FIELD_REF offsets.
17929         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
17930         wi::to_poly_offset.
17931         * var-tracking.c (emit_note_insn_var_location): Use
17932         tree_to_poly_uint64.
17934 2018-05-29  Jim Wilson  <jimw@sifive.com>
17936         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
17938 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
17940         PR target/85950
17941         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
17942         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
17943         sequence.
17944         (sse4_1_round<mode>2): Use nonimmediate_operand
17945         for operand 1 predicate.
17947 2018-05-29  Martin Sebor  <msebor@redhat.com>
17948             Richard Biener  <rguenther@suse.de>
17950         PR testsuite/85888
17951         * calls.c (get_size_range): Call determine_value_range instead
17952         of get_value_range..
17953         * tree-vrp.h (determine_value_range): Declared new function.
17954         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
17956 2018-05-29  Richard Biener  <rguenther@suse.de>
17958         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
17959         sure to use non-pattern stmts for get_earlier_stmt arguments.
17960         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
17961         called on pattern stmts.
17962         (get_later_stmt): Likewise.
17964 2018-05-29  Martin Liska  <mliska@suse.cz>
17966         PR gcov-profile/85759
17967         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
17968         env variables.
17970 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
17972         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
17973         VEC_UNPACK_*_EXPR.
17974         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
17975         VEC_PACK_*_EXPR.
17977         PR target/85918
17978         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
17979         VEC_PACK_FLOAT_EXPR): New tree codes.
17980         * tree-pretty-print.c (op_code_prio): Handle
17981         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
17982         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
17983         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
17984         * tree-inline.c (estimate_operator_cost): Likewise.
17985         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
17986         * fold-const.c (const_binop): Likewise.
17987         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
17988         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
17989         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
17990         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
17991         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
17992         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
17993         * expr.c (expand_expr_real_2): Likewise.
17994         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
17995         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
17996         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
17997         optabs.
17998         * optabs.c (expand_widen_pattern_expr): For
17999         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
18000         sign from result type rather than operand's type.
18001         (expand_binop_directly): For vec_packu_float_optab and
18002         vec_packs_float_optab allow result type to be different from operand's
18003         type.
18004         * optabs-tree.c (optab_for_tree_code): Handle
18005         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
18006         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
18007         * tree-vect-generic.c (expand_vector_operations_1):  Handle
18008         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
18009         VEC_PACK_FLOAT_EXPR.
18010         * tree-vect-stmts.c (supportable_widening_operation): Handle
18011         FIX_TRUNC_EXPR.
18012         (supportable_narrowing_operation): Handle FLOAT_EXPR.
18013         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
18014         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
18015         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
18016         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
18017         mode attributes.
18018         (vec_pack<floatprefix>_float_<mode>): New expander.
18019         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
18020         attributes.
18021         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
18022         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
18023         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
18024         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
18025         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
18026         Document.
18027         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
18028         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
18029         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
18030         VEC_PACK_FLOAT_EXPR): Document.
18032 2018-05-29  Richard Biener  <rguenther@suse.de>
18034         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
18035         member.
18036         (stmt_vec_info_vec): Make pointer.
18037         (init_stmt_vec_info_vec): Remove.
18038         (free_stmt_vec_info_vec): Likewise.
18039         (set_stmt_vec_info_vec): New function.
18040         (free_stmt_vec_infos): Likewise.
18041         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
18042         (set_vinfo_for_stmt): Likewise.
18043         (get_earlier_stmt): Likewise.
18044         (get_later_stmt): Likewise.
18045         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
18046         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
18047         (vec_info::~vec_info): Free stmt_vec_infos.
18048         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
18049         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
18050         (pass_slp_vectorize::execute): Likewise.
18051         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
18052         (free_stmt_vec_info_vec): Likewise.
18053         (set_stmt_vec_info_vec): New function.
18054         (free_stmt_vec_infos): Likewise.
18055         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
18056         the global stmt_vec_info_vec.
18057         * tree-parloops.c (gather_scalar_reductions): Use
18058         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
18059         vector.
18061 2018-05-29  Richard Biener  <rguenther@suse.de>
18063         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
18065 2018-05-29  Martin Liska  <mliska@suse.cz>
18066             David Malcolm  <dmalcolm@redhat.com>
18068         * vec.c (test_reverse): New.
18069         (vec_c_tests): Add new test.
18070         * vec.h (vl_ptr>::reverse): New function.
18072 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
18074         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
18076         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
18077         and later.
18079 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
18081         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
18083 2018-05-28  Richard Biener  <rguenther@suse.de>
18085         PR tree-optimization/85933
18086         * tree-vect-data-refs.c (vect_record_base_alignments): Only
18087         look at stmts marked as vectorizable.
18089 2018-05-28  Richard Biener  <rguenther@suse.de>
18091         PR tree-optimization/85934
18092         * tree-vect-generic.c (expand_vector_operations_1): Hoist
18093         vector boolean check before scalar optimization.
18095 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
18097         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
18098         for armv5te.
18100 2018-05-28  Mark Wielaard  <mark@klomp.org>
18102         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
18103         if it is an expression containing a minus sign.
18105 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
18107         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
18109 2018-05-27  Paul Koning  <ni1d@arrl.net>
18111         * config/pdp11/pdp11.md (truncsihi2): Remove.
18113 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
18114             Chung-Ju Wu  <jasonwucj@gmail.com>
18116         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
18117         implementation.
18118         (unaligned_store_dw): Ditto.
18119         * config/nds32/nds32-memory-manipulation.c
18120         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
18121         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
18122         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
18123         (emit_setmem_word_loop): Rename to ...
18124         (emit_setmem_doubleword_loop): ... this.
18125         (nds32_gen_dup_4_byte_to_word_value): New function.
18126         (nds32_gen_dup_8_byte_to_double_word_value): New function.
18127         (nds32_expand_setmem_loop): Refine implementation.
18128         (nds32_expand_setmem_loop_v3m): Ditto.
18129         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
18130         pattern.
18132 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
18134         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
18136 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
18138         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
18139         (nds32_init_machine_status): Initialize machine->attr_naked_p and
18140         machine->attr_no_prologue_p.
18141         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
18142         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
18143         (nds32_expand_epilogue): Consider attr_naked_p.
18144         (nds32_expand_epilogue_v3pop): Likewise.
18145         (nds32_can_use_return_insn): Likewise.
18146         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
18147         attr_no_prologue_p fields.
18148         * config/nds32/nds32.opt (mret-in-naked-func): New option.
18150 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
18152         PR target/85918
18153         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
18154         attributes.
18155         * config/i386/sse.md
18156         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
18157         Rename to ...
18158         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
18159         ... this.
18160         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
18161         Rename to ...
18162         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
18163         ... this.
18164         (*<floatsuffix>floatv2div2sf2): Rename to ...
18165         (*float<floatunssuffix>v2div2sf2): ... this.
18166         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
18167         (float<floatunssuffix>v2div2sf2_mask): ... this.
18168         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
18169         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
18170         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
18171         to ...
18172         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
18173         ... this.
18174         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
18175         Rename to ...
18176         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
18177         ... this.
18178         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
18179         Rename to ...
18180         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
18181         ... this.
18182         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
18183         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
18184         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
18185         gen_ufix_truncv8dfv8si2.
18186         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
18187         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
18188         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
18189         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
18190         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
18191         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
18192         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
18193         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
18195 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
18197         PR target/85900
18198         PR target/85345
18199         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
18201 2018-05-25  Jim Wilson  <jimw@sifive.com>
18203         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
18204         * config/riscv/riscv.c (struct machine_function): Add
18205         interrupt_handler_p and attribute_checked_p fields.
18206         (riscv_attribute_table): Add interrupt.
18207         (riscv_interrupt_type_p): New.
18208         (riscv_save_reg_p): Save extra regs for interrupt handler.
18209         (riscv_use_save_libcall): Return false  for interrupt handler.
18210         (riscv_first_stack_step): Add forward declaration.
18211         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
18212         for interrupt handler with large frame.  Use it for saved reg list.
18213         (riscv_expand_prologue): Move flag_stack_usage_info support to
18214         eliminate duplication.
18215         (riscv_expand_epilogue): Generate mret for interrupt handler.
18216         (riscv_epilogue_uses): New.
18217         (riscv_can_use_return_insn): Return false for interrupt handler.
18218         (riscv_function_ok_for_sibcall): Likewise.
18219         (riscv_set_current_function): Add interrupt handler support.
18220         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
18221         * config/riscv/riscv.md (UNSPECV_MRET): New.
18222         (GP_REGNUM): New.
18223         (riscv_frflags, riscv_fsflags): Use tab after opcode.
18224         (riscv_mret): New.
18225         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
18227 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
18229         PR tree-optimization/85712
18230         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
18231         this candidate has already been replaced in-situ by a copy.
18233 2018-05-25  Jason Merrill  <jason@redhat.com>
18235         PR c++/80485 - inline function non-zero address.
18236         * symtab.c (nonzero_address): Check DECL_COMDAT.
18238 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
18240         PR target/83628
18241         * config/alpha/alpha.md (ashlsi3): New insn pattern.
18242         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
18243         extension of SImode operation.  Use const123_operand predicate.
18244         (*saddsi_1): Remove.
18245         (*saddl_se_1): Ditto.
18246         (*ssubsi_1): Ditto.
18247         (*ssubl_se_1): Ditto.
18248         * config/alpha/predicates.md (const123_operand): New predicate.
18249         * config/alpha/constraints.md (P): Use IN_RANGE.
18251 2018-05-25  Richard Biener  <rguenther@suse.de>
18253         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
18254         defaulted to true.
18255         (ref_maybe_used_by_stmt_p): Likewise.
18256         (stmt_may_clobber_ref_p): Likewise.
18257         (stmt_may_clobber_ref_p_1): Likewise.
18258         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
18259         and pass it along.
18260         (ref_maybe_used_by_stmt_p): Likewise.
18261         (stmt_may_clobber_ref_p): Likewise.
18262         (stmt_may_clobber_ref_p_1): Likewise.
18263         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
18264         the alias oracle to disambiguate DRs with stmts DR analysis
18265         couldn't handle.
18266         (vect_analyze_data_refs): Do not give up on not analyzable
18267         DRs for BB vectorization.  Remove code truncating the dataref
18268         vector.
18270 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
18272         PR target/85832
18273         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
18274         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
18275         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
18277 2018-05-25  Richard Biener  <rguenther@suse.de>
18279         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
18280         function, combining stmt data ref gathering and fatal analysis
18281         parts.
18282         (vect_analyze_data_refs): Remove now redudnant code and simplify.
18283         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
18284         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
18285         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
18286         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
18288 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
18290         PR tree-optimization/85720
18291         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
18292         SCC if all partitions are builtins.
18293         (version_loop_by_alias_check): New parameter.  Generate cancelable
18294         runtime alias check if all partitions are builtins.
18295         (distribute_loop): Update call to above function.
18297 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
18299         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
18300         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
18301         (parm_default_def_partition_arg): Ditto.
18302         (set_parm_default_def_partition): Ditto.
18303         (get_parm_default_def_partitions): Ditto and make it static.
18304         (get_undefined_value_partitions): Ditto and make it static.
18305         (remove_ssa_form): Refactor call to init_var_map here.
18306         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
18307         computation for loop region.
18308         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
18309         (register_default_def): Delete.
18310         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
18311         (parm_default_def_partition_arg): Ditto.
18312         (set_parm_default_def_partition): Ditto.
18313         (get_parm_default_def_partitions): Ditto and make it static.
18314         (get_undefined_value_partitions): Ditto and make it static.
18315         (coalesce_with_default, coalesce_with_default): Update comment.
18316         (create_coalesce_list_for_region): New func factored out from
18317         create_outofssa_var_map.
18318         (populate_coalesce_list_for_outofssa): New func factored out from
18319         create_outofssa_var_map and coalesce_ssa_name.
18320         (create_outofssa_var_map): Delete.
18321         (coalesce_ssa_name): Refactor to support live range computation.
18322         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
18323         (get_parm_default_def_partitions): Delete.
18324         (get_undefined_value_partitions): Ditto.
18325         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
18326         computation for loop region.
18327         (new_tree_live_info, loe_visit_block): Ditto.
18328         (live_worklist, set_var_live_on_entry): Ditto.
18329         (calculate_live_on_exit, verify_live_on_entry): Ditto.
18330         * tree-ssa-live.h (struct _var_map): New fields.
18331         (init_var_map): Change decl.
18332         (region_contains_p): New.
18334 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
18336         * tree-ssa-live.h (live_merge_and_clear): Delete.
18338 2018-05-25  Richard Biener  <rguenther@suse.de>
18340         PR c++/85912
18341         * tree-dump.c (dequeue_and_dump): Remove access to removed
18342         operand 2 of a SWITCH_EXPR.
18344 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18346         * doc/sourcebuild.texi (vect_double_cond_arith): Include
18347         multiplication and division.
18348         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
18349         (cond_udiv@var{m}, cond_umod@var{m}): Document.
18350         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
18351         (cond_udiv_optab, cond_umod_optab): New optabs.
18352         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
18353         (IFN_COND_RDIV): New internal functions.
18354         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
18355         TRUNC_MOD_EXPR and RDIV_EXPR.
18356         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
18357         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
18358         New unspecs.
18359         (SVE_INT_BINARY): Include mult.
18360         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
18361         (optab, sve_int_op): Handle mult.
18362         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
18363         UNSPEC_COND_DIV.
18364         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
18365         for SVE_INT_BINARY_SD.
18367 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18369         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
18370         (optab, sve_int_op): Handle div and udiv.
18371         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
18372         for SVE_INT_BINARY_SD.
18373         (*<optab><mode>3): New insn for the same.
18375 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18377         * tree-vect-patterns.c: Include predict.h.
18378         (vect_recog_divmod_pattern): Restrict check for division support
18379         to when optimizing for size.
18381 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18383         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
18384         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
18385         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
18386         (gimple_match_op::set_op): Likewise.
18387         (gimple_resimplify4): Declare.
18388         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
18389         (expr::gen_transform): Likewise.
18390         (decision_tree::gen): Generate a simplification routine for 4 operands.
18391         * gimple-match-head.c (gimple_simplify): Add an overload for
18392         4 operands.  In the top-level function, handle up to 4 call
18393         arguments and call gimple_resimplify4.
18394         (gimple_resimplify4): New function.
18395         (build_call_internal): Pass a fourth operand.
18396         (maybe_push_to_seq): Likewise.
18397         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
18398         Fold VEC_COND_EXPRs of an operation and a default value into
18399         an IFN_COND_* function if possible.
18400         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
18401         New unspecs.
18402         (SVE_COND_FP_BINARY): Include them.
18403         (optab, sve_fp_op): Handle them.
18404         (SVE_INT_BINARY_REV): New code iterator.
18405         (SVE_COND_FP_BINARY_REV): New int iterator.
18406         (commutative): New int attribute.
18407         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
18408         Declare.
18409         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
18410         function.
18411         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
18412         (*cond_<optab><mode>): New patterns for reversed operands.
18414 2018-05-25  Richard Biener  <rguenther@suse.de>
18416         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
18417         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
18418         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
18419         (STMT_VINFO_GROUPED_ACCESS): Adjust.
18420         * tree-vect-data-refs.c (everywhere): Adjust users.
18421         * tree-vect-loop.c (everywhere): Likewise.
18422         * tree-vect-slp.c (everywhere): Likewise.
18423         * tree-vect-stmts.c (everywhere): Likewise.
18424         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
18426 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
18428         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
18429         Rename to...
18430         (gcc_cv_as_section_exclude): ... this.
18431         Try Solaris as #exclude syntax.
18432         * configure: Regenerate.
18433         * config.in: Regenerate.
18434         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
18435         SECTION_EXCLUDE.
18436         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
18437         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
18439         * varasm.c (default_elf_asm_named_section): Don't check if
18440         HAVE_GAS_SECTION_EXCLUDE is defined.
18442 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18444         * doc/md.texi: Update the documentation of the cond_* optabs
18445         to mention the new final operand.  Fix GET_MODE_NUNITS call.
18446         Describe the scalar case too.
18447         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
18448         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
18449         instead of 2.
18450         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
18451         (get_conditional_internal_fn): Update comment.
18452         * tree-vect-loop.c (vectorizable_reduction): Pass the original
18453         accumulator value as a final argument to conditional functions.
18454         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
18455         a define_expand and add an "else" operand.  Assert for now that
18456         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
18457         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
18458         (*cond_<optab><mode>): New patterns.
18459         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
18460         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
18461         (UNSPEC_COND_EOR): Delete.
18462         (optab): Remove associated mappings.
18463         (SVE_INT_BINARY): New code iterator.
18464         (sve_int_op): Remove int attribute and add "minus" to the code
18465         attribute.
18466         (SVE_COND_INT_OP): Delete.
18467         (SVE_COND_FP_OP): Rename to...
18468         (SVE_COND_FP_BINARY): ...this.
18470 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18472         * optabs.c (can_reuse_operands_p): New function.
18473         (maybe_legitimize_operands): Try to reuse the results for
18474         earlier operands.
18476 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
18478         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
18479         Add {q} suffix to insn mnemonic.
18481 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
18483         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
18484         (msp430_warn_func_return): New.
18486 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
18488         * fold-const.c (tree_nonzero_bits): New function.
18489         * fold-const.h (tree_nonzero_bits): Likewise.
18490         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
18491         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
18493 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
18495         PR target/85900
18496         PR target/85345
18497         * varasm.c (assemble_alias): Check ifunc_resolver only on
18498         FUNCTION_DECL.
18500 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
18502         PR target/85903
18503         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
18504         when memory input operand is handled.
18506 2018-05-24  Luis Machado  <luis.machado@linaro.org>
18508         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
18509         global.
18510         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
18512 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
18514         * match.pd: Delay FMA folds until after vectorization.
18516 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
18518         PR target/83009
18519         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
18520         address check not strict.
18522 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
18524         * gimple-match.h (gimple_match_op): New class.
18525         (mprts_hook): Replace parameters with a gimple_match_op *.
18526         (maybe_build_generic_op): Likewise.
18527         (gimple_simplified_result_is_gimple_val): Replace parameters with
18528         a const gimple_match_op *.
18529         (gimple_simplify): Replace code_helper * and tree * parameters with
18530         a gimple_match_op * parameter.
18531         (gimple_resimplify1): Replace code_helper *, tree and tree *
18532         parameters with a gimple_match_op * parameter.
18533         (gimple_resimplify2): Likewise.
18534         (gimple_resimplify3): Likewise.
18535         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
18536         parameters with a gimple_match_op * parameter.
18537         * gimple-match-head.c (gimple_simplify): Change prototypes of
18538         auto-generated functions to take a gimple_match_op * instead of
18539         separate code_helper * and tree * parameters.  Make the same
18540         change in the top-level overload and update calls to the
18541         gimple_resimplify routines.  Update calls to the auto-generated
18542         functions and to maybe_push_res_to_seq in the publicly-facing
18543         operation-specific gimple_simplify overloads.
18544         (gimple_match_op::MAX_NUM_OPS): Define.
18545         (gimple_resimplify1): Replace rcode and ops with a single res_op
18546         parameter.  Update call to gimple_simplify.
18547         (gimple_resimplify2): Likewise.
18548         (gimple_resimplify3): Likewise.
18549         (mprts_hook): Replace parameters with a gimple_match_op *.
18550         (maybe_build_generic_op): Likewise.
18551         (build_call_internal): Replace type, nargs and ops with
18552         a gimple_match_op *.
18553         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
18554         with a single gimple_match_op *.  Update calls to mprts_hook,
18555         build_call_internal and gimple_simplified_result_is_gimple_val.
18556         Factor out code that is common to the tree_code and combined_fn cases.
18557         * genmatch.c (expr::gen_transform): Replace tem_code and
18558         tem_ops with a gimple_match_op called tem_op.  Update calls
18559         to the gimple_resimplify functions and maybe_push_res_to_seq.
18560         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
18561         res_ops.  Update call to the gimple_resimplify functions.
18562         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
18563         (decision_tree::gen): Make the functions take a gimple_match_op *
18564         called res_op instead of separate res_code and res_ops parameters.
18565         Update call accordingly.
18566         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
18567         and ops with a single res_op parameter.  Update calls to
18568         maybe_build_generic_op and maybe_push_res_to_seq.
18569         (fold_stmt_1): Update calls to gimple_simplify and
18570         replace_stmt_with_simplification.
18571         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
18572         and gimple_simplified_result_is_gimple_val.
18573         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
18574         gimple_simplify.
18575         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
18576         with a gimple_match_op *.
18577         (vn_nary_build_or_lookup): Likewise.  Update call to
18578         vn_nary_build_or_lookup_1.
18579         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
18580         gimple_match_op *.  Update calls to the gimple_resimplify routines
18581         and to gimple_simplified_result_is_gimple_val.
18582         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
18583         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
18584         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
18585         (visit_nary_op): Likewise.
18586         (visit_reference_op_load): Likewise.
18588 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18590         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
18591         modifier for printing the step amount.
18593 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
18595         PR target/78849
18596         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
18597         types.
18599 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
18601         * doc/sourcebuild.texi (Endianness): New subsubsection.
18603 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18605         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18606         <prefetch_dynamic_strides>: New const bool field.
18607         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18608         prefetch_dynamic_strides.
18609         (exynosm1_prefetch_tune): Likewise.
18610         (thunderxt88_prefetch_tune): Likewise.
18611         (thunderx_prefetch_tune): Likewise.
18612         (thunderx2t99_prefetch_tune): Likewise.
18613         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
18614         false.
18615         (aarch64_override_options_internal): Update to set
18616         PARAM_PREFETCH_DYNAMIC_STRIDES.
18617         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
18618         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
18619         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
18620         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
18621         prefetch-dynamic-strides setting.
18623 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18625         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18626         <minimum_stride>: New const int field.
18627         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18628         minimum_stride field defaulting to -1.
18629         (exynosm1_prefetch_tune): Likewise.
18630         (thunderxt88_prefetch_tune): Likewise.
18631         (thunderx_prefetch_tune): Likewise.
18632         (thunderx2t99_prefetch_tune): Likewise.
18633         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
18634         <default_opt_level>: Set to 3.
18635         (aarch64_override_options_internal): Update to set
18636         PARAM_PREFETCH_MINIMUM_STRIDE.
18637         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
18638         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
18639         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
18640         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
18641         stride is constant and is below the minimum stride threshold.
18643 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18645         * config/arm/arm-cpus.in (mode26): Delete.
18646         (armv4): Delete mode26 reference.
18647         * config/arm/arm.c (arm_configure_build_target): Delete use of
18648         isa_bit_mode26.
18650 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
18652         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
18653         New insn pattern.
18654         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
18655         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
18656         for non-SSE modes.
18657         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
18658         (floatunsdidf2): Ditto.
18660 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
18662         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
18663         (fixuns_trunc<mode>si2_avx512f): Ditto.
18664         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
18665         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
18666         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
18668 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
18670         PR rtl-optimization/79985
18671         * df-scan.c (df_insn_refs_collect): Remove special case for
18672         global registers and asm statements.
18674 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
18676         * extend.texi (Global Register Variables): Rewrite the bullet list.
18677         Note that the register is available for allocation. Note that access
18678         via inline asm must use constraints. Add note about async-signal
18679         handlers. Remove paragraph about automagic register selection.
18681 2018-05-23  Richard Biener  <rguenther@suse.de>
18683         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
18684         of fixed offset from memset VN.
18686 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
18688         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
18689         first_interp field.
18690         (alloc_cand_and_find_basis): Initialize first_interp field.
18691         (slsr_process_mul): Modify first_interp field.
18692         (slsr_process_add): Likewise.
18693         (slsr_process_cast): Modify first_interp field for each new
18694         interpretation.
18695         (slsr_process_copy): Likewise.
18696         (dump_candidate): Dump first_interp field.
18697         (replace_mult_candidate): Process all interpretations, not just
18698         subsequent ones.
18699         (replace_rhs_if_not_dup): Likewise.
18700         (replace_one_candidate): Likewise.
18702 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
18704         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
18705         Add new boolean.
18706         (aarch64_needs_frame_chain): New function.
18707         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
18709 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
18711         PR target/84882
18712         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
18713         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
18714         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
18715         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
18716         as true for strict-align.
18717         (aarch64_can_inline_p): Perform checks even when callee has no
18718         attributes to check for strict alignment.
18719         * doc/extend.texi (AArch64 Function Attributes): Document
18720         no-strict-align.
18721         * doc/invoke.texi: (AArch64 Options): Likewise.
18723 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
18725         PR tree-optimization/85853
18726         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
18727         the handling of the root of the node to...
18728         (vect_slp_analyze_node_operations_1): ...this new function,
18729         and run the whole thing with the child nodes' def types
18730         set according to their SLP node's def type.
18732 2018-05-23  Richard Biener  <rguenther@suse.de>
18734         PR middle-end/85874
18735         * tree-data-ref.c (create_runtime_alias_checks): Defer
18736         and ignore overflow warnings.
18738 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
18740         PR tree-optimization/85822
18741         * tree-vrp.c (is_masked_range_test): Fix handling of negative
18742         constants.
18744 2018-05-23  Richard Biener  <rguenther@suse.de>
18746         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
18747         memset constants via native_interpret_expr.
18749 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
18751         PR target/85345
18752         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
18753         attribute.
18754         (cgraph_node::create_alias): Likewise.
18755         (cgraph_node::get_availability): Check ifunc_resolver instead
18756         of looking up ifunc attribute.
18757         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
18758         * varasm.c (do_assemble_alias): Likewise.
18759         (assemble_alias): Likewise.
18760         (default_binds_local_p_3): Likewise.
18761         * cgraph.h (cgraph_node): Add ifunc_resolver.
18762         (cgraph_node::only_called_directly_or_aliased_p): Return false
18763         for IFUNC resolver.
18764         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
18765         attribute.
18766         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
18767         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
18768         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
18769         instead of looking up ifunc attribute.
18771 2018-05-22  Luis Machado  <luis.machado@linaro.org>
18773         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
18775 2018-05-22  Martin Sebor  <msebor@redhat.com>
18777         PR middle-end/85359
18778         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
18779         only when expasion succeeds.
18780         (expand_builtin_strcmp): Same.
18781         (expand_builtin_strncmp): Same.
18783 2018-05-22  Martin Sebor  <msebor@redhat.com>
18785         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
18787 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
18788             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18790         * config/aarch64/aarch64-ldpstp.md: Replace uses of
18791         aarch64_mem_pair_operand with memory_operand and delete operand swapping
18792         code.
18793         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
18794         Add check for legitimate_address.
18795         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
18796         (aarch64_swap_ldrstr_operands): New.
18797         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
18798         Define prototype.
18800 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
18801             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18803         * config/aarch64/aarch64.md: New patterns to generate stp
18804         and ldp.
18805         (store_pair_sw, store_pair_dw): New patterns to generate stp for
18806         single words and double words.
18807         (load_pair_sw, load_pair_dw): Likewise.
18808         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
18809         Delete.
18810         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
18811         Delete.
18812         * config/aarch64/aarch64-ldpstp.md: Modify peephole
18813         for different mode ldpstp and add peephole for merged zero stores.
18814         Likewise for loads.
18815         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
18816         Add size check.
18817         (aarch64_gen_store_pair): Rename calls to match new patterns.
18818         (aarch64_gen_load_pair): Rename calls to match new patterns.
18819         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
18820         (load_pair<DREG:mode><DREG2:mode>): ... This.
18821         (store_pair<mode>): Rename to...
18822         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
18823         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
18824         New mode iterators.
18825         (V_INT_EQUIV): Handle SImode.
18826         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
18827         New predicate.
18829 2018-05-22  Martin Sebor  <msebor@redhat.com>
18831         PR c/85623
18832         * calls.c (maybe_warn_nonstring_arg): Use string length to set
18833         or ajust the presumed bound on an operation to avoid unnecessary
18834         warnings.
18836 2018-05-22  Martin Sebor  <msebor@redhat.com>
18838         PR tree-optimization/85826
18839         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
18840         assuming that a DECL necesarily has a constant size.
18842 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
18844         PR middle-end/85862
18845         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
18847 2018-05-22  Richard Biener  <rguenther@suse.de>
18849         PR tree-optimization/85834
18850         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
18851         non-constant and non-zero memset arguments.
18853 2018-05-22  Martin Liska  <mliska@suse.cz>
18855         PR ipa/85607
18856         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
18858 2018-05-22  Richard Biener  <rguenther@suse.de>
18860         PR tree-optimization/85863
18861         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
18862         comparisons when vectype is specified.
18863         (vectorizable_condition): Do not specify vectype for
18864         vect_is_simple_cond when SLP vectorizing.
18866 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
18868         PR target/85657
18869         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
18870         define __ibm128 as long double.
18871         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
18872         as a distinct type when IEEE 128-bit support is enabled.
18873         (init_float128_ieee): Fix up conversions between IFmode and IEEE
18874         128-bit types to use the correct functions.
18875         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
18876         convert between 128-bit floating point types that have different
18877         modes but the same representation, instead of using gen_lowpart to
18878         makean alias.
18879         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
18880         KFmode.
18881         (IFKF_reg): New attributes to give the register constraints for
18882         IFmode and KFmode.
18883         (extend<mode>tf2_internal): New insns to mark an explicit
18884         conversion between 128-bit floating point types that have a
18885         different mode but share the same representation.
18887 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
18889         PR tree-optimization/85814
18890         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
18891         a null return from get_strinfo when unsharing the next
18892         strinfo in the chain.
18894 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
18896         PR gcc/84923
18897         * varasm.c (weak_finish): Clean up weak_decls.
18899 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18901         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
18902         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
18903         UNSPEC_UADALP values.
18904         * config/aarch64/iterators.md (ABAL): New int iterator.
18905         (ABDL2): Likewise.
18906         (ADALP): Likewise.
18907         (sur): Add mappings for the above.
18908         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
18909         New define_insn.
18910         (aarch64_<sur>abal<mode>_4): Likewise.
18911         (aarch64_<sur>adalp<mode>_3): Likewise.
18912         (<sur>sadv16qi): New define_expand.
18914 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
18916         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
18917         (*movdf_internal): Ditto.
18918         (*rcpsf2_sse): Ditto.
18919         (*rsqrtsf2_sse): Ditto.
18920         (*sqrt<mode>2_sse): Ditto.
18922 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
18924         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
18925         eor3q<mode>4.
18926         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
18927         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
18928         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
18929         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
18930         vbcaxq_s64): New.
18931         * config/aarch64/arm_neon.h: Likewise.
18932         * config/aarch64/iterators.md (VQ_I): New.
18934 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
18936         * config.gcc: Add arc/t-multilib-linux to tmake_file for
18937         arc*-*-linux*.
18938         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
18939         MULTILIB_DIRNAMES
18941 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
18943         * config/nds32/constraints.md (S): New constraint.
18944         * config/nds32/nds32.md (call_internal): Use constraint S.
18945         (call_value_internal): Likewise.
18946         (sibcall_internal): Likewise.
18947         (sibcall_value_internal): Likewise.
18949 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
18950             Chung-Ju Wu  <jasonwucj@gmail.com>
18952         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
18953         into consideration.
18955 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
18956             Chung-Ju Wu  <jasonwucj@gmail.com>
18958         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
18959         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
18960         (nds32_rtx_costs_impl): Simplify.
18961         (nds32_address_cost_impl): Simplify.
18962         (nds32_init_rtx_costs): New function.
18963         (nds32_rtx_costs_speed_prefer): Likewise.
18964         (nds32_rtx_costs_size_prefer): Likewise.
18965         (nds32_address_cost_speed_prefer): Likewise.
18966         (nds32_address_cost_speed_fwprop): Likewise.
18967         (nds32_address_cost_size_prefer): Likewise.
18968         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
18969         * config/nds32/nds32.c (nds32_option_override): Use
18970         nds32_init_rtx_costs function.
18972 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
18974         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
18975         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
18976         (TARGET_PIPELINE_N8): Likewise.
18977         (TARGET_PIPELINE_N10): Likewise.
18978         (TARGET_PIPELINE_N13): Likewise.
18979         (TARGET_PIPELINE_GRAYWOLF): Likewise.
18981 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
18983         * config/nds32/nds32-fpu.md: Update copyright year.
18985 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18987         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
18989 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18991         * config/nds32/nds32.c
18992         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
18993         * config/nds32/nds32.opt (minline-asm-r15): New option.
18995 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
18997         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
18998         MASK_HW_ABS.
18999         * config/nds32/nds32.md (abssi2): New pattern.
19001 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
19003         * config/i386/i386.md (rex64namesuffix): New mode attribute.
19004         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
19005         Merge insn pattern from sse_cvtsi2ss<round_name> and
19006         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
19007         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
19008         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
19009         using SWI48 mode iterator.
19010         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
19011         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
19012         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
19013         pattern from sse_cvttss2si<round_saeonly_name>
19014         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
19015         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
19016         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
19017         using SWI48 mode iterator.
19018         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
19019         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
19020         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
19021         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
19022         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
19023         using SWI48 mode iterator.
19024         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
19025         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
19026         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
19027         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
19028         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
19029         SWI48 mode iterator.
19030         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
19031         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
19032         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
19033         pattern from sse_cvttsd2si<round_saeonly_name>
19034         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
19036 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19038         * config/nds32/nds32-md-auxiliary.c
19039         (nds32_valid_smw_lwm_base_p): Refine.
19040         (nds32_output_smw_single_word): Refine.
19041         (nds32_output_smw_double_word): New.
19042         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
19044 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19046         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
19047         (nds32_output_stack_pop): Refine.
19048         (nds32_expand_unaligned_load): Refine.
19049         (nds32_expand_unaligned_store): Refine.
19051 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
19052             Chung-Ju Wu  <jasonwucj@gmail.com>
19054         * config/nds32/constants.md: Add TP_REGNUM constant.
19055         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
19056         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
19057         UNSPEC_ADD32.
19058         * config/nds32/nds32-doubleword.md: Consider flag_pic.
19059         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
19060         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
19061         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
19062         and PIC code generation.
19063         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
19064         code generation.
19065         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
19066         optimization.
19067         * config/nds32/nds32.md: Support TLS and PIC.
19068         * config/nds32/nds32.c: Support TLS and PIC.
19069         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
19070         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
19071         predicate.
19073 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19075         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
19076         mode with E_ prefix.
19078 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
19079             Chung-Ju Wu  <jasonwucj@gmail.com>
19081         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
19082         * config/nds32/nds32-md-auxiliary.c
19083         (symbolic_reference_mentioned_p): New.
19084         (nds32_legitimize_ict_address): New.
19085         (nds32_expand_ict_move): New.
19086         (nds32_indirect_call_referenced_p): New.
19087         (nds32_symbol_binds_local_p): Delete.
19088         (nds32_long_call_p): Modify.
19089         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
19090         * config/nds32/nds32-protos.h
19091         (symbolic_reference_mentioned_p): Declare.
19092         (nds32_legitimize_ict_address): Declare.
19093         (nds32_expand_ict_move): Declare.
19094         (nds32_indirect_call_referenced_p): Declare.
19095         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
19096         (nds32_relax_group): Use nds32_ict_const_p as condition.
19097         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
19098         (nds32_asm_file_start): Output ict_model directive in asm code.
19099         (nds32_legitimate_address_p): Consider indirect call.
19100         (nds32_print_operand): Consider indirect call.
19101         (nds32_print_operand_address): Consider indirect call.
19102         (nds32_insert_attributes): Handle "indirect_call" attribute.
19103         (TARGET_LEGITIMATE_ADDRESS_P): Define.
19104         (TARGET_LEGITIMATE_CONSTANT_P): Define.
19105         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
19106         (TARGET_DELEGITIMIZE_ADDRESS): Define.
19107         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
19108         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
19109         (TARGET_ICT_MODEL_SMALL): Define.
19110         (TARGET_ICT_MODEL_LARGE): Define.
19111         * config/nds32/nds32.md (movsi): Consider ict model.
19112         (call, call_value): Consider ict model.
19113         (sibcall, sibcall_value): Consider ict model.
19114         * config/nds32/nds32.opt (mict-model): New option.
19115         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
19116         model.
19118 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
19119             Monk Chiang  <sh.chiang04@gmail.com>
19120             Jim Wilson <jimw@sifive.com>
19122         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
19123         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
19124         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
19125         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
19126         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
19127         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
19128         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
19129         compute save_libcall_adjustment properly.
19130         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
19131         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
19132         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
19133         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
19134         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
19135         (ABI_SPEC): Handle mabi=ilp32e.
19136         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
19137         (RVE): Add RVE mask.
19138         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
19139         <-march>: Add rv32e as an example.
19141 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
19143         PR c++/82899
19144         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
19145         (intra_create_variable_infos): Handle C++ constructors.
19147 2018-05-18  Martin Liska  <mliska@suse.cz>
19149         * passes.def: Remove a redundant pass.
19151 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
19153         PR bootstrap/85838
19154         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
19156 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19158         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
19159         (ARMv4): Update.
19160         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
19161         (ARMv6m): Update.
19162         (armv2, armv2a, armv3, armv3m): Delete architectures.
19163         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
19164         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
19165         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
19166         Delete cpus.
19167         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
19168         (*mulsidi3adddi): Likewise.
19169         (mulsidi3): Likewise.
19170         (*mulsidi3_nov6): Likewise.
19171         (umulsidi3): Likewise.
19172         (umulsidi3_nov6): Likewise.
19173         (umaddsidi4): Likewise.
19174         (*umulsidi3adddi): Likewise.
19175         (smulsi3_highpart): Likewise.
19176         (*smulsi3_highpart_nov6): Likewise.
19177         (umulsi3_highpart): Likewise.
19178         (*umulsi3_highpart_nov6): Likewise.
19179         * config/arm/arm.h (arm_arch3m): Delete.
19180         * config/arm/arm.c (arm_arch3m): Delete.
19181         (arm_option_override_internal): Update armv3-related comment.
19182         (arm_configure_build_target): Delete use of isa_bit_mode32.
19183         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
19184         (arm_rtx_costs_internal): Delete check of arm_arch3m.
19185         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
19186         (mulsa3): Likewise.
19187         (mulusa3): Likewise.
19188         * config/arm/arm-protos.h (arm_arch3m): Delete.
19189         * config/arm/arm-tables.opt: Regenerate.
19190         * config/arm/arm-tune.md: Likewise.
19191         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
19192         deleted architectures.
19194 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19196         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
19197         (armv5t, armv5te): New features.
19198         (ARMv5, ARMv5e): Delete fgroups.
19199         (ARMv5t, ARMv5te): Adjust for above changes.
19200         (ARMv6m): Likewise.
19201         (armv5, armv5e): Delete arches.
19202         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
19203         arm_arch5.
19204         (*call_reg_arm): Likewise.
19205         (*call_value_reg_armv5): Likewise.
19206         (*call_value_reg_arm): Likewise.
19207         (*call_symbol): Likewise.
19208         (*call_value_symbol): Likewise.
19209         (*sibcall_insn): Likewise.
19210         (*sibcall_value_insn): Likewise.
19211         (clzsi2): Likewise.
19212         (prefetch): Likewise.
19213         (define_split and define_peephole2 dependent on arm_arch5):
19214         Likewise.
19215         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
19216         arm_arch5e.
19217         (TARGET_ARM_QBIT): Likewise.
19218         (TARGET_DSP_MULTIPLY): Likewise.
19219         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
19220         (arm_arch5, arm_arch5e): Delete.
19221         (arm_arch5t, arm_arch5te): Declare.
19222         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
19223         (arm_arch5t): Declare.
19224         (arm_option_reconfigure_globals): Update for the above.
19225         (arm_options_perform_arch_sanity_checks): Update comment, replace
19226         use of arm_arch5 with arm_arch5t.
19227         (use_return_insn): Likewise.
19228         (arm_emit_call_insn): Likewise.
19229         (output_return_instruction): Likewise.
19230         (arm_final_prescan_insn): Likewise.
19231         (arm_coproc_builtin_available): Likewise.
19232         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
19233         arm_arch5e with arm_arch5t and arm_arch5te.
19234         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
19235         (arm_arch5t, arm_arch5te): Declare.
19236         * config/arm/arm-tables.opt: Regenerate.
19237         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
19238         * config/arm/t-multilib: Likewise.
19239         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
19240         instead of arm_arch5.
19241         (*call_reg_thumb1): Likewise.
19242         (*call_value_reg_thumb1_v5): Likewise.
19243         (*call_value_reg_thumb1): Likewise.
19244         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
19245         unreachable path.
19246         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
19248 2018-05-18  Martin Liska  <mliska@suse.cz>
19250         PR gcov-profile/84846
19251         * doc/gcov.texi: Document -t option of gcov tool.
19253 2018-05-18  Martin Liska  <mliska@suse.cz>
19255         PR gcov-profile/84846
19256         * gcov.c (print_usage): Add new -t option.
19257         (process_args): Handle the option.
19258         (generate_results): Use stdout as output when requested by
19259         the option.
19261 2018-05-18  Martin Liska  <mliska@suse.cz>
19263         PR gcov-profile/84846
19264         * coverage.c (coverage_init): Write PWD to .gcno file.
19265         * doc/gcov.texi: Document how working directory is printed.
19266         * gcov-dump.c (dump_gcov_file): Print PWD.
19267         * gcov.c (output_intermediate_file): Likewise.
19268         (read_graph_file): Read PWD string.
19269         (output_lines): Print PWD.
19271 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19273         PR middle-end/85817
19274         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
19275         for retval and return false if all args to phi are zero.
19277 2018-05-18  Richard Biener  <rguenther@suse.de>
19279         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
19280         method.
19281         (evrp_dom_walker::before_dom_children): Call it.
19283 2018-05-18  Richard Biener  <rguenther@suse.de>
19285         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
19286         results when processing array refs with variable index.
19288 2018-05-18  Toon Moene  <toon@moene.org>
19290         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
19291         directly after that of -floop-interchange. Indicate that both
19292         options are enabled by default when specifying -O3.
19294 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19296         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
19297         iterator.  Delete separate integer-mode vec_set<mode> expander.
19298         (aarch64_simd_vec_setv2di): Delete.
19299         (vec_setv2di): Delete.
19300         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
19301         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
19302         the "w, r" alternative.
19304 2018-05-18  Martin Liska  <mliska@suse.cz>
19306         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
19307         * tree-pass.h (make_pass_lower_switch_O0): New function.
19308         * tree-switch-conversion.c (node_has_low_bound): Remove.
19309         (node_has_high_bound): Likewise.
19310         (node_is_bounded): Likewise.
19311         (class pass_lower_switch): Make it a template type and create
19312         two instances.
19313         (pass_lower_switch::execute): Add template argument.
19314         (make_pass_lower_switch): New function.
19315         (make_pass_lower_switch_O0): New function.
19316         (do_jump_if_equal): Remove.
19317         (emit_case_nodes): Simplify to just handle all 3 cases and leave
19318         all the hard work to tree optimization passes.
19320 2018-05-18  Martin Liska  <mliska@suse.cz>
19322         * dbgcnt.c (limit_low): Renamed from limit.
19323         (limit_high): New variable.
19324         (dbg_cnt_is_enabled): Check for upper limit.
19325         (dbg_cnt): Adjust dumping.
19326         (dbg_cnt_set_limit_by_index): Add new argument for high
19327         value.
19328         (dbg_cnt_set_limit_by_name): Likewise.
19329         (dbg_cnt_process_single_pair): Parse new format.
19330         (dbg_cnt_process_opt): Use strtok.
19331         (dbg_cnt_list_all_counters): Remove 'value' and add
19332         'limit_high'.
19333         * doc/invoke.texi: Document changes.
19335 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
19337         * doc/sourcebuild.texi (scalar_all_fma): Document.
19338         * tree.def (FMA_EXPR): Delete.
19339         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
19340         * internal-fn.c (ternary_direct): New macro.
19341         (expand_ternary_optab_fn): Likewise.
19342         (direct_ternary_optab_supported_p): Likewise.
19343         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
19344         * builtins.c (fold_builtin_fma): Delete.
19345         (fold_builtin_3): Don't call it.
19346         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
19347         * expr.c (expand_expr_real_2): Likewise.
19348         * fold-const.c (operand_equal_p): Likewise.
19349         (fold_ternary_loc): Likewise.
19350         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
19351         * gimple.c (DEFTREECODE): Likewise.
19352         * gimplify.c (gimplify_expr): Likewise.
19353         * optabs-tree.c (optab_for_tree_code): Likewise.
19354         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
19355         * tree-eh.c (operation_could_trap_p): Likewise.
19356         (stmt_could_throw_1_p): Likewise.
19357         * tree-inline.c (estimate_operator_cost): Likewise.
19358         * tree-pretty-print.c (dump_generic_node): Likewise.
19359         (op_code_prio): Likewise.
19360         * tree-ssa-loop-im.c (stmt_cost): Likewise.
19361         * tree-ssa-operands.c (get_expr_operands): Likewise.
19362         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
19363         * fold-const-call.h (fold_fma): Delete.
19364         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
19365         CFN_FNMA and CFN_FNMS.
19366         (fold_fma): Delete.
19367         * genmatch.c (combined_fn): New enum.
19368         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
19369         (commutative_op): New function.
19370         (commutate): Use it.  Handle more than 2 operands.
19371         (dt_operand::gen_gimple_expr): Use commutative_op.
19372         (parser::parse_expr): Allow :c to be used with non-binary
19373         operators if the commutative operand is known.
19374         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
19375         CFN_FMS, CFN_FNMA and CFN_FNMS.
19376         (backprop::process_assign_use): Remove FMA_EXPR handling.
19377         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
19378         (gen_hsa_fma): New function.
19379         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
19380         IFN_FNMA and IFN_FNMS.
19381         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
19382         * gimple-fold.h (follow_all_ssa_edges): Declare.
19383         * gimple-fold.c (follow_all_ssa_edges): New function.
19384         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
19385         gimple_build interface and use follow_all_ssa_edges to fold the result.
19386         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
19387         instead of checking for optabs directly.
19388         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
19389         rather than FMA_EXPRs.
19390         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
19391         call to IFN_FMA instead of an FMA_EXPR.
19393 2018-05-17  Jim Wilson  <jimw@sifive.com>
19395         * expr.c (do_tablejump): When converting index to Pmode, if we have a
19396         sign extended promoted subreg, and the range does not have the sign bit
19397         set, then do a sign extend.
19399         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
19400         test, check for sign extended subreg and/or constant operands, and
19401         do a sign extend in that case.
19403 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
19405         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
19406         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
19407         Add untyped.
19408         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
19409         Change logics_shift_reg to logics_shift_imm.
19410         (thunderx2t99_fp_loadpair_basic): Delete.
19411         (thunderx2t99_fp_storepair_basic): Delete.
19412         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
19413         (thunderx2t99_asimd_polynomial): Delete.
19414         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
19415         and neon_fp_mul_d_scalar_q.
19416         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
19417         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
19418         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
19419         (thunderx2t99_asimd_lut): Add missing tbl types.
19420         (thunderx2t99_asimd_ext): Delete.
19421         (thunderx2t99_asimd_load1_1_mult): Delete.
19422         (thunderx2t99_asimd_load1_2_mult): Delete.
19423         (thunderx2t99_asimd_load1_ldp): New.
19424         (thunderx2t99_asimd_load1): New.
19425         (thunderx2t99_asimd_load2): Add missing *load2* types.
19426         (thunderx2t99_asimd_load3): New.
19427         (thunderx2t99_asimd_load4): New.
19428         (thunderx2t99_asimd_store1_1_mult): Delete.
19429         (thunderx2t99_asimd_store1_2_mult): Delete.
19430         (thunderx2t99_asimd_store2_mult): Delete.
19431         (thunderx2t99_asimd_store2_onelane): Delete.
19432         (thunderx2t99_asimd_store_stp): New.
19433         (thunderx2t99_asimd_store1): New.
19434         (thunderx2t99_asimd_store2): New.
19435         (thunderx2t99_asimd_store3): New.
19436         (thunderx2t99_asimd_store4): New.
19438 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
19440         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
19441         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
19443 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
19444             Segher Boessenkool  <segher@kernel.crashing.org>
19446         PR target/85698
19447         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
19448         operand.
19450 2018-05-17  Richard Biener  <rguenther@suse.de>
19452         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
19453         for pruning loop and prune defs feeding only already visited PHIs.
19455 2018-05-17  Richard Biener  <rguenther@suse.de>
19457         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
19459 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
19460             Richard Biener  <rguenther@suse.de>
19462         PR tree-optimization/85793
19463         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
19464         for VMAT_ELEMENTWISE.
19466 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
19468         * internal-fn.h (lookup_internal_fn): Declare
19469         * internal-fn.c (lookup_internal_fn): New function.
19470         * gimple.c (gimple_build_call_from_tree): Handle calls to
19471         internal functions.
19472         * gimple-pretty-print.c (dump_gimple_call): Print "." before
19473         internal function names.
19474         * tree-pretty-print.c (dump_generic_node): Likewise.
19475         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
19477 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
19479         * gimple-fold.h (gimple_build): Make the function forms take
19480         combined_fn rather than built_in_function.
19481         (gimple_simplify): Likewise.
19482         * gimple-match-head.c (gimple_simplify): Likewise.
19483         * gimple-fold.c (gimple_build): Likewise.
19484         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
19485         rather than gimple_build_call_internal.
19486         (get_initial_defs_for_reduction): Likewise.
19487         (vect_create_epilog_for_reduction): Likewise.
19488         (vectorizable_live_operation): Likewise.
19490 2018-05-17  Martin Liska  <mliska@suse.cz>
19492         * gimple-ssa-sprintf.c (format_directive): Do not use
19493         space in between 'G_' and '('.
19495 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
19497         PR target/85323
19498         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
19499         even if the mask is not all ones.
19501         PR target/85323
19502         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
19503         vector.
19504         (ix86_gimple_fold_builtin): Likewise.
19506         PR target/85323
19507         * config/i386/i386.c: Include tree-vector-builder.h.
19508         (ix86_vector_shift_count): New function.
19509         (ix86_fold_builtin): Fold shift builtins by scalar count.
19510         (ix86_gimple_fold_builtin): Likewise.
19512         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
19513         _mm512_setzero): New intrinsics.
19515 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
19516             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19518         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
19519         code generation for cases where splatting a value is not useful.
19520         * simplify-rtx.c (simplify_ternary_operation): Simplify
19521         vec_merge across a vec_duplicate and a paradoxical subreg forming
19522         a vector mode to a vec_concat.
19524 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
19526         * config.gcc: Support "goldmont-plus".
19527         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
19528         "goldmont-plus".
19529         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19530         PROCESSOR_GOLDMONT_PLUS.
19531         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
19532         (processor_target_table): Add "goldmont-plus".
19533         (PTA_GOLDMONT_PLUS): Define.
19534         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
19535         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
19536         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
19537         (fold_builtin_cpu): Add "goldmont-plus".
19538         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
19539         (ix86_option_override_internal): Add "goldmont-plus".
19540         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
19541         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
19542         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
19543         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
19545 2018-05-17  Richard Biener  <rguenther@suse.de>
19547         PR tree-optimization/85757
19548         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
19549         remove defs that only feed that PHI from further processing.
19551 2018-05-16  Jim Wilson  <jimw@sifive.com>
19553         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
19554         asterisk to name.
19555         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
19557 2018-05-16  Mark Wielaard  <mark@klomp.org>
19559         * dwarf2out.c (count_index_strings): New function.
19560         (output_indirect_strings): Call count_index_strings and generate
19561         header for dwarf_version >= 5.
19563 2018-05-16  Mark Wielaard  <mark@klomp.org>
19565         * dwarf2out.c (dwarf_FORM): New function.
19566         (set_indirect_string): Use dwarf_FORM.
19567         (reset_indirect_string): Likewise.
19568         (size_of_die): Likewise.
19569         (value_format): Likewise.
19570         (output_die): Likewise.
19571         (add_skeleton_AT_string): Likewise.
19572         (output_macinfo_op): Likewise.
19573         (index_string): Likewise.
19574         (output_index_string_offset): Likewise.
19575         (output_index_string): Likewise.
19576         (count_index_strings): Likewise.
19578 2018-05-16  Carl Love  <cel@us.ibm.com>
19580         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
19581         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
19583 2018-05-16  Martin Jambor  <mjambor@suse.cz>
19585         * ipa-prop.c (ipa_free_all_edge_args): Remove.
19586         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
19588 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
19590         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
19591         (fnma<mode>4): Likewise.
19592         (fms<mode>4): Likewise.
19593         (fnms<mode>4): Likewise.
19594         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
19595         (aarch64_fnma<mode>4): Likewise.
19596         (aarch64_fms<mode>4): Likewise.
19597         (aarch64_fnms<mode>4): Likewise.
19598         (aarch64_fnmadd<mode>4): Likewise.
19600 2018-05-16  Jason Merrill  <jason@redhat.com>
19602         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
19604 2018-05-16  Richard Biener  <rguenther@suse.de>
19606         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
19607         (dump_stmt_cost): Declare.
19608         (add_stmt_cost): Dump cost we add.
19609         (add_stmt_costs): New function.
19610         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
19611         No longer exported.
19612         (vect_analyze_stmt): Adjust prototype.
19613         (vectorizable_condition): Likewise.
19614         (vectorizable_live_operation): Likewise.
19615         (vectorizable_reduction): Likewise.
19616         (vectorizable_induction): Likewise.
19617         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
19618         cost vector to pass to vectorizable_ and record afterwards.
19619         (vect_model_reduction_cost): Take cost vector argument and adjust.
19620         (vect_model_induction_cost): Likewise.
19621         (vectorizable_reduction): Likewise.
19622         (vectorizable_induction): Likewise.
19623         (vectorizable_live_operation): Likewise.
19624         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
19625         SLP_TREE_NUMBER_OF_VEC_STMTS.
19626         (vect_analyze_slp_cost_1): Remove.
19627         (vect_analyze_slp_cost): Likewise.
19628         (vect_slp_analyze_node_operations): Take visited args and
19629         a target cost vector.  Avoid processing already visited stmt sets.
19630         (vect_slp_analyze_operations): Use a local cost vector to gather
19631         costs and register those of non-discarded instances.
19632         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
19633         (vect_schedule_slp_instance): Remove copying of
19634         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
19635         zero.
19636         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
19637         adding cost.  Record cost entry location.
19638         (vect_prologue_cost_for_slp_op): Function to compute cost of
19639         a constant or invariant generated for SLP vect in the prologue,
19640         split out from vect_analyze_slp_cost_1.
19641         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
19642         (vect_model_promotion_demotion_cost): Likewise.
19643         (vect_model_store_cost): Likewise, make static.
19644         (vect_model_load_cost): Likewise.
19645         (vectorizable_bswap): Add cost vector arg and adjust.
19646         (vectorizable_call): Likewise.
19647         (vectorizable_simd_clone_call): Likewise.
19648         (vectorizable_conversion): Likewise.
19649         (vectorizable_assignment): Likewise.
19650         (vectorizable_shift): Likewise.
19651         (vectorizable_operation): Likewise.
19652         (vectorizable_store): Likewise.
19653         (vectorizable_load): Likewise.
19654         (vectorizable_condition): Likewise.
19655         (vectorizable_comparison): Likewise.
19656         (can_vectorize_live_stmts): Likewise.
19657         (vect_analyze_stmt): Likewise.
19658         (vect_transform_stmt): Adjust calls to vectorizable_*.
19659         * tree-vectorizer.c: Include gimple-pretty-print.h.
19660         (dump_stmt_cost): New function.
19662 2018-05-16  Richard Biener  <rguenther@suse.de>
19664         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
19665         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
19666         * tree-ssa-dse.c: Include tree-ssa-loop.h.
19667         (check_name): New callback.
19668         (dse_classify_store): Track cycles via a visited bitmap of PHI
19669         defs and simplify handling of in-loop and across loop dead stores
19670         and properly fail for loop-variant refs.  Handle byte-tracking with
19671         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
19672         limiting the walk.
19674 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
19676         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
19677         (vect_get_mask_type_for_stmt): Likewise.
19678         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
19679         split out from...
19680         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
19681         to determine the statement's vector type and the vector type that
19682         should be used for calculating nunits.  Deal with cases in which
19683         the type has to be deferred.
19684         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
19685         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
19686         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
19687         (vect_determine_vf_for_stmt): New functions, split out from...
19688         (vect_determine_vectorization_factor): ...here.
19689         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
19690         (vect_get_mask_type_for_stmt): New functions, split out from
19691         vect_determine_vectorization_factor.
19693 2018-05-16  Richard Biener  <rguenther@suse.de>
19695         * tree-cfg.c (verify_gimple_assign_ternary): Properly
19696         verify the [VEC_]COND_EXPR embedded comparison.
19698 2018-05-15  Martin Sebor  <msebor@redhat.com>
19700         PR tree-optimization/85753
19701         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
19702         RECORD_TYPE in addition to ARRAY_TYPE.
19704 2018-05-15  Martin Sebor  <msebor@redhat.com>
19706         PR middle-end/85643
19707         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
19709 2018-05-15  Richard Biener  <rguenther@suse.de>
19711         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
19712         add by_clobber_p one.  Change algorithm to collect all defs
19713         representing uses we need to walk and try reducing them to
19714         a single one before failing.
19715         (dse_dom_walker::dse_optimize_stmt): Adjust.
19717 2018-05-13  Mark Wielaard  <mark@klomp.org>
19719         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
19720         (size_of_loc_descr): Likewise.
19721         (output_loc_operands): Likewise.
19722         (output_loc_operands_raw): Likewise.
19723         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
19724         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
19725         (hash_loc_operands): Likewise.
19726         (compare_loc_operands): Likewise.
19728 2018-05-14  Mark Wielaard  <mark@klomp.org>
19730         * dwarf2out.c (count_index_addrs): New function.
19731         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
19733 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19735         PR tree-optimization/83648
19736         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
19737         return value as malloc candidate.
19739 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19741         PR ipa/85734
19742         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
19743         param as true in call to suggest_attribute.
19745 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
19747         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
19748         -mreadonly-in-sdata.
19750 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19752         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
19753         New pattern.
19754         (aarch64_crypto_aesd_fused): Likewise.
19756 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
19758         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
19759         (movsi_aarch64): Likewise.
19760         (load_pairsi): Likewise.
19761         (load_pairdi): Likewise.
19762         (store_pairsi): Likewise.
19763         (store_pairdi): Likewise.
19764         (load_pairsf): Likewise.
19765         (load_pairdf): Likewise.
19766         (store_pairsf): Likewise.
19767         (store_pairdf): Likewise.
19768         (zero_extend): Likewise.
19769         (trunc): Swap alternatives.
19770         (fcvt_target): Add '?' to prefer w over r.
19772 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
19774         PR target/85756
19775         * config/i386/i386.md: Disallow non-commutative arithmetics in
19776         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
19777         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
19778         in the peephole2 before it.
19780 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
19782         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
19783         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
19784         (ix86_handle_option): Handle -mcldemote.
19785         * config.gcc: New header.
19786         * config/i386/cldemoteintrin.h: New file.
19787         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
19788         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
19789         -mcldemote.
19790         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19791         OPTION_MASK_ISA_CLDEMOTE.
19792         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
19793         (ix86_valid_target_attribute_inner_p): Ditto.
19794         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
19795         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
19796         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
19797         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
19798         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
19799         (cldemote): New.
19800         * config/i386/i386.opt: Add -mcldemote.
19801         * config/i386/x86intrin.h: New header.
19802         * doc/invoke.texi: Add -mcldemote.
19804 2018-05-14  Richard Biener  <rguenther@suse.de>
19806         * doc/match-and-simplify.texi: Adjust :s documentation.
19808 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
19810         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
19811         intended memcpy size.
19812         (REORDER_45): Likewise.
19814 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
19816         * sort.cc: New file.
19817         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
19818         * vec.c (qsort_chk): Use gcc_qsort.
19819         * Makefile.in (OBJS-libcommon): Add sort.o.
19820         (build/sort.o): New target.  Use it...
19821         (BUILD_RTL): ... here, and...
19822         (build/gencfn-macros): ... here, and...
19823         (build/genmatch): ... here.
19825 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19826             Chung-Ju Wu  <jasonwucj@gmail.com>
19828         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
19829         * config/nds32/nds32-graywolf.md: New file.
19830         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
19831         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
19832         pipeline.
19833         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
19834         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
19835         * config/nds32/nds32.md (pipeline_model): Add graywolf.
19836         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
19837         * config/nds32/pipelines.md: Include n15 settings.
19839 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19840             Chung-Ju Wu  <jasonwucj@gmail.com>
19842         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
19843         * config/nds32/nds32-n13.md: New file.
19844         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
19845         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
19846         pipeline.
19847         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
19848         * config/nds32/nds32.md (pipeline_model): Add n13.
19849         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
19850         * config/nds32/pipelines.md: Include n13 settings.
19852 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19853             Chung-Ju Wu  <jasonwucj@gmail.com>
19855         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
19856         * config/nds32/nds32-n10.md: New file.
19857         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
19858         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
19859         pipeline.
19860         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
19861         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
19862         * config/nds32/nds32.md (pipeline_model): Add n10.
19863         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
19864         * config/nds32/pipelines.md: Include n10 settings.
19866 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
19867             Kito Cheng  <kito.cheng@gmail.com>
19868             Chung-Ju Wu  <jasonwucj@gmail.com>
19870         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
19871         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
19872         Add enum values for DSP extension instructions.
19873         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
19874         New constraints.
19875         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
19876         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
19877         New code iterators.
19878         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
19879         * config/nds32/nds32-dspext.md: New file for DSP implementation.
19880         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
19881         * config/nds32/nds32-intrinsic.md: Likewise.
19882         * config/nds32/nds32_intrinsic.h: Likewise.
19883         * config/nds32/nds32-md-auxiliary.c: Likewise.
19884         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
19885         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
19886         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
19887         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
19888         * config/nds32/nds32-protos.h: New declarations for DSP extension.
19889         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
19890         TYPE_DMAC in switch statement.
19891         * config/nds32/nds32.c: New checking and implementation for DSP
19892         extension instructions.
19893         * config/nds32/nds32.h: Likewise.
19894         * config/nds32/nds32.md: Likewise.
19895         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
19896         * config/nds32/predicates.md: Implement new predicates for DSP
19897         extension.
19899 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
19901         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
19902         Reformat alternatives and attributes so it is easier to identify
19903         which constraints/attributes go with which instruction.
19904         (mov<mode>_hardfloat32, FMOVE64): Likewise.
19905         (mov<mode>_softfloat32, FMOVE64): Likewise.
19906         (mov<mode>_hardfloat64, FMOVE64): Likewise.
19907         (mov<mode>_softfloat64, FMOVE64): Likewise.
19909 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19911         * doc/extend.texi (PowerPC Built-in Functions): Rename this
19912         subsection.
19913         (Basic PowerPC Built-in Functions): The new name of the
19914         subsection previously known as "PowerPC Built-in Functions".
19915         (Basic PowerPC Built-in Functions Available on all Configurations):
19916         New subsubsection.
19917         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
19918         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
19919         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
19920         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
19922 2018-05-11  Martin Jambor  <mjambor@suse.cz>
19924         PR ipa/85655
19925         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
19926         single const.
19928 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
19930         PR target/85733
19931         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
19933 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
19935         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
19936         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
19937         (ix86_handle_option): Handle -mwaitpkg.
19938         * config.gcc: New header.
19939         * config/i386/cpuid.h (bit_WAITPKG): New bit.
19940         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
19941         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
19942         function type.
19943         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19944         OPTION_MASK_ISA_WAITPKG.
19945         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
19946         (ix86_option_override_internal): Add PTA_WAITPKG.
19947         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
19948         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
19949         IX86_BUILTIN_TPAUSE.
19950         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
19951         __builtin_ia32_umwait and __builtin_ia32_tpause.
19952         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
19953         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
19954         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
19955         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
19956         UNSPECV_TPAUSE): New.
19957         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
19958         * config/i386/i386.opt: Add -mwaitpkg.
19959         * config/i386/waitpkgintrin.h: New file.
19960         * config/i386/x86intrin.h: New header.
19961         * doc/invoke.texi: Add -mwaitpkg.
19963 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
19965         PR target/85606
19966         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
19967         equivalent.
19968         (cortex-m0): Use armv6s-m isa.
19969         (cortex-m0plus): Likewise.
19970         (cortex-m1): Likewise.
19971         (cortex-m0.small-multiply): Likewise.
19972         (cortex-m0plus.small-multiply): Likewise.
19973         (cortex-m1.small-multiply): Likewise.
19975 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
19976             Jakub Jelinek  <jakub@redhat.com>
19978         PR tree-optimization/85692
19979         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
19980         source permute as well.
19982 2018-05-11  Martin Liska  <mliska@suse.cz>
19984         PR sanitizer/85556
19985         * doc/extend.texi: Document LLVM style format for no_sanitize
19986         attribute.
19988 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
19990         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
19991         mode_supports_vsx_dform_quad to mode_supports_dq_form.
19992         (mode_supports_vsx_dform_quad): Likewise.
19993         (mode_supports_vmx_dform): Move these functions to be next to the
19994         other mode_supports functions.
19995         (mode_supports_dq_form): Likewise.
19996         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
19997         mode_supports_dq_form.
19998         (reg_offset_addressing_ok_p): Likewise.
19999         (offsettable_ok_by_alignment): Likewise.
20000         (rs6000_legitimate_offset_address_p): Likewise.
20001         (legitimate_lo_sum_address_p): Likewise.
20002         (rs6000_legitimize_address): Likewise.
20003         (rs6000_legitimize_reload_address): Likewise.
20004         (rs6000_secondary_reload_inner): Likewise.
20005         (rs6000_preferred_reload_class): Likewise.
20006         (rs6000_output_move_128bit): Likewise.
20008 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
20010         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
20011         Generate SImode target register for null target.
20012         <case IX86_BUILTIN_XGETBV>: Ditto.
20013         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
20014         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
20016 2018-05-10  Carl Love  <cel@us.ibm.com>
20018         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
20019         dcbtt and dcbtstt if operands[2] is 0.
20021 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
20023         PR target/85693
20024         * config/i386/sse.md (usadv64qi): New expander.
20026 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
20028         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
20029         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
20030         -maltivec=be support.
20031         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
20032         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
20033         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
20034         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
20035         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
20036         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
20037         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
20038         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
20039         altivec_vsumsws): Adjust.
20040         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
20041         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
20042         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
20043         support.
20044         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
20045         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
20046         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
20047         (altivec_lve<VI_char>x): Delete expand.
20048         (*altivec_lve<VI_char>x_internal): Rename to...
20049         (altivec_lve<VI_char>x): ... this.
20050         (altivec_lvxl_<mode>): Delete expand.
20051         (*altivec_lvxl_<mode>_internal): Rename to ...
20052         (altivec_lvxl_<mode>): ... this.
20053         (altivec_stvxl_<mode>): Delete expand.
20054         (*altivec_stvxl_<mode>_internal): Rename to ...
20055         (altivec_stvxl_<mode>): ... this.
20056         (altivec_stve<VI_char>x): Delete expand.
20057         (*altivec_stve<VI_char>x_internal): Rename to ...
20058         (altivec_stve<VI_char>x): ... this.
20059         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
20060         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
20061         reduc_plus_scal_<mode>): Adjust.
20062         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
20063         comment.
20064         (rs6000_cpu_cpp_builtins): Adjust.
20065         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
20066         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
20067         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
20068         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
20069         -maltivec=be support.
20070         (rs6000_split_vec_extract_var): Adjust.
20071         (rs6000_split_v4si_init): Adjust.
20072         (swap_selector_for_mode): Delete.
20073         (altivec_expand_lvx_be, altivec_expand_stvx_be,
20074         altivec_expand_stvex_be): Delete.
20075         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
20076         -maltivec=be support.
20077         (rs6000_gimple_fold_builtin): Ditto.
20078         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
20079         Adjust.
20080         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
20081         (TARGET_DIRECT_MOVE_64BIT): Adjust.
20082         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
20083         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
20084         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
20085         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
20086         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
20087         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
20088         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
20089         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
20090         anonymous split): Adjust.
20091         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
20092         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
20094 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
20096         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
20097         when --with-gxx-include-dir is also specified.
20098         * configure: Regenerate.
20100 2018-05-09  Jim Wilson  <jimw@sifive.com>
20102         PR target/84797
20103         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
20104         * config/riscv/t-withmultilib: New.
20105         * config/riscv/withmultilib.h: New.
20106         * doc/install.texi: Document RISC-V --with-multilib-list support.
20108 2018-05-09  Richard Biener  <rguenther@suse.de>
20110         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
20111         vector.
20112         (vect_bb_vectorization_profitable_p): Adjust.  Compute
20113         actual scalar cost using the cost vector and the add_stmt_cost
20114         machinery.
20116 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
20118         PR rtl-optimization/85645
20119         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
20120         in the REG_CFA_REGISTER note for LR, don't leave it empty.
20122 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
20124         PR rtl-optimization/85645
20125         * shrink-wrap.c (spread_components): Return a boolean saying if
20126         anything was changed.
20127         (try_shrink_wrapping_separate): Iterate spread_components until
20128         nothing changes anymore.
20130 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
20132         PR rtl-optimization/85645
20133         * regrename.c (build_def_use): Also kill the chains that include the
20134         destination of a REG_CFA_REGISTER note.
20136 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
20138         PR rtl-optimization/85645
20139         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
20140         insn that has a REG_CFA_REGISTER note.
20142 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
20144         * cfgexpand.c (expand_clobber): New function.
20145         (expand_gimple_stmt_1): Use it.
20146         * tree-vect-stmts.c (vect_clobber_variable): New function,
20147         split out from...
20148         (vectorizable_simd_clone_call): ...here.
20149         (vectorizable_store): Emit a clobber either side of an
20150         IFN_STORE_LANES sequence.
20151         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
20153 2018-05-09  Tom de Vries  <tom@codesourcery.com>
20155         PR target/85626
20156         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
20157         (define_insn "trap_if_false"): Add exit after trap.
20159 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
20161         PR rtl-optimization/85638
20162         * bb-reorder.c: Include common/common-target.h.
20163         (create_forwarder_block): New function extracted from...
20164         (fix_up_crossing_landing_pad): ...here.  Rename into...
20165         (dw2_fix_up_crossing_landing_pad): ...this.
20166         (sjlj_fix_up_crossing_landing_pad): New function.
20167         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
20168         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
20169         from both partitions and exit the loop after one iteration.
20171 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20173         Revert:
20174         * doc/extend.texi (PowerPC Built-in Functions): Rename this
20175         subsection.
20176         (Basic PowerPC Built-in Functions): The new name of the
20177         subsection previously known as "PowerPC Built-in Functions".
20178         (Basic PowerPC Built-in Functions Available on all Configurations):
20179         New subsubsection.
20180         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
20181         subsubsection.
20182         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
20183         subsubsection.
20184         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
20185         subsubsection.
20186         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
20187         subsubsection.
20189 2018-05-08  Jim Wilson  <jimw@sifive.com>
20191         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
20192         (LD_EMUL_SUFFIX): New.
20193         (LINK_SPEC): Use it.
20195 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20197         * doc/extend.texi (PowerPC Built-in Functions): Rename this
20198         subsection.
20199         (Basic PowerPC Built-in Functions): The new name of the
20200         subsection previously known as "PowerPC Built-in Functions".
20201         (Basic PowerPC Built-in Functions Available on all Configurations):
20202         New subsubsection.
20203         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
20204         subsubsection.
20205         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
20206         subsubsection.
20207         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
20208         subsubsection.
20209         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
20210         subsubsection.
20212 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
20214         PR target/85683
20215         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
20216         after cmpelim optimization.
20218 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
20220         * config.gcc: Support "goldmont".
20221         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
20222         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20223         PROCESSOR_GOLDMONT.
20224         * config/i386/i386.c (m_GOLDMONT): Define.
20225         (processor_target_table): Add "goldmont".
20226         (PTA_GOLDMONT): Define.
20227         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
20228         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
20229         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
20230         (fold_builtin_cpu): Add "goldmont".
20231         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
20232         (ix86_option_override_internal): Add "goldmont".
20233         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
20234         (processor_type): Add PROCESSOR_GOLDMONT.
20235         * config/i386/i386.md: Add CPU "glm".
20236         * config/i386/glm.md: New file.
20237         * config/i386/x86-tune.def: Add m_GOLDMONT.
20238         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
20240 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
20242         PR target/85572
20243         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
20244         E_V4DImode.
20245         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
20246         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
20247         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
20249         PR target/85317
20250         * config/i386/i386.c (ix86_fold_builtin): Handle
20251         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
20253         PR target/85480
20254         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
20255         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
20257 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
20259         PR target/85658
20260         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
20261         (check_arch): Likewise.
20262         (check_fpu): Return the result rather than printing it.
20263         (end arch): Fix operator precedence.
20264         (end cpu): Likewise.
20265         (END): Print the result from check_fpu.
20267 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
20268             Alan Hayward  <alan.hayward@arm.com>
20269             David Sherwood  <david.sherwood@arm.com>
20271         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
20272         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
20273         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
20274         (*fcmuo<mode>_and): New patterns.
20276 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
20278         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
20279         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
20280         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
20281         (cmp_op, sve_imm_con): New code attributes.
20282         (SVE_COND_INT_CMP, imm_con): Delete.
20283         (cmp_op): Remove above unspecs from int attribute.
20284         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
20285         to...
20286         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
20287         comparison-specific unspecs.
20288         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
20289         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
20290         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
20291         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
20292         (*vec_fcm<cmp_op><mode>): Rename to...
20293         (*fcm<cmp_op><mode>): ...this and adjust likewise.
20294         (*vec_fcmuo<mode>): Rename to...
20295         (*fcmuo<mode>): ...this and adjust likewise.
20296         (*pred_fcm<cmp_op><mode>): New pattern.
20297         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
20298         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
20299         functions.
20300         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
20301         and UNORDERED.
20302         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
20303         (aarch64_emit_sve_predicated_cond): New function.
20304         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
20305         (aarch64_emit_unspec_cond_or): Replace with...
20306         (aarch64_emit_sve_or_conds): ...this new function.  Use
20307         aarch64_emit_sve_ptrue_op for the individual comparisons and
20308         aarch64_emit_binop to OR them together.
20309         (aarch64_emit_inverted_unspec_cond): Replace with...
20310         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
20311         aarch64_emit_sve_ptrue_op for the comparison and
20312         aarch64_emit_unop to invert the result.
20313         (aarch64_expand_sve_vec_cmp_float): Update after the above
20314         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
20316 2018-05-07  Nathan Sidwell  <nathan@acm.org>
20318         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
20319         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
20320         (Backwards Compatibility): Likewise.
20322 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20324         PR bootstrap/85681
20325         Revert:
20326         2018-05-07  Luis Machado  <luis.machado@linaro.org>
20328         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20329         <prefetch_dynamic_strides>: New const bool field.
20330         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20331         prefetch_dynamic_strides.
20332         (exynosm1_prefetch_tune): Likewise.
20333         (thunderxt88_prefetch_tune): Likewise.
20334         (thunderx_prefetch_tune): Likewise.
20335         (thunderx2t99_prefetch_tune): Likewise.
20336         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
20337         to false.
20338         (aarch64_override_options_internal): Update to set
20339         PARAM_PREFETCH_DYNAMIC_STRIDES.
20340         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
20341         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
20342         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
20343         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
20344         prefetch-dynamic-strides setting.
20346         2018-05-07  Luis Machado  <luis.machado@linaro.org>
20348         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20349         <minimum_stride>: New const int field.
20350         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20351         minimum_stride field.
20352         (exynosm1_prefetch_tune): Likewise.
20353         (thunderxt88_prefetch_tune): Likewise.
20354         (thunderx_prefetch_tune): Likewise.
20355         (thunderx2t99_prefetch_tune): Likewise.
20356         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
20357         (aarch64_override_options_internal): Update to set
20358         PARAM_PREFETCH_MINIMUM_STRIDE.
20359         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
20360         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
20361         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
20362         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
20363         stride is constant and is below the minimum stride threshold.
20365 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20367         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
20368         to 512.
20370 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20372         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20373         <prefetch_dynamic_strides>: New const bool field.
20374         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20375         prefetch_dynamic_strides.
20376         (exynosm1_prefetch_tune): Likewise.
20377         (thunderxt88_prefetch_tune): Likewise.
20378         (thunderx_prefetch_tune): Likewise.
20379         (thunderx2t99_prefetch_tune): Likewise.
20380         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
20381         to false.
20382         (aarch64_override_options_internal): Update to set
20383         PARAM_PREFETCH_DYNAMIC_STRIDES.
20384         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
20385         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
20386         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
20387         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
20388         prefetch-dynamic-strides setting.
20390 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20392         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20393         <minimum_stride>: New const int field.
20394         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20395         minimum_stride field.
20396         (exynosm1_prefetch_tune): Likewise.
20397         (thunderxt88_prefetch_tune): Likewise.
20398         (thunderx_prefetch_tune): Likewise.
20399         (thunderx2t99_prefetch_tune): Likewise.
20400         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
20401         (aarch64_override_options_internal): Update to set
20402         PARAM_PREFETCH_MINIMUM_STRIDE.
20403         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
20404         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
20405         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
20406         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
20407         stride is constant and is below the minimum stride threshold.
20409 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
20411         PR c++/85659
20412         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
20413         the type is addressable.  Don't force op into register if it has
20414         BLKmode.
20416 2018-05-05  Roland McGrath  <mcgrathr@google.com>
20418         PR other/77609
20419         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
20420         any section for which we don't know a specific type it should have,
20421         regardless of name.  Previously this was done only for the exact
20422         names ".init_array", ".fini_array", and ".preinit_array".
20423         (default_elf_asm_named_section): Add comment about
20424         relationship with default_section_type_flags and SECTION_NOTYPE.
20425         (get_section): Don't consider it a type conflict if one side has
20426         SECTION_NOTYPE and the other doesn't, as long as neither has the
20427         SECTION_BSS et al used in the default_section_type_flags logic.
20429 2018-05-05  Tom de Vries  <tom@codesourcery.com>
20431         PR target/85653
20432         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
20433         (workaround_barsyncs): New function.
20434         (nvptx_reorg): Use workaround_barsyncs.
20435         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
20436         (define_expand "nvptx_membar_cta"): New define_expand.
20437         (define_insn "*nvptx_membar_cta"): New insn.
20439 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
20441         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
20442         To improve optimization opportunities.
20443         * builtin-types.def: The new needed builtin types for the above.
20445 2018-05-04  Richard Biener  <rguenther@suse.de>
20447         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
20448         * gimple-ssa-store-merging.c
20449         (imm_store_chain_info::output_merged_store): Remove redundant create,
20450         release split_store vector contents on failure.
20451         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
20452         scalar stmt vector on cache hit.
20454 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
20456         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
20457         Xilinx FP support.
20458         * config.gcc (powerpc-xilinx-eabi*): Remove.
20459         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
20460         support.
20461         (fusion_addis_mem_combo_load): Ditto.
20462         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
20463         FP support.
20464         (rs6000_cpu_cpp_builtins): Ditto.
20465         * config/rs6000/rs6000-linux.c
20466         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
20467         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
20468         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
20469         support.
20470         (rs6000_setup_reg_addr_masks): Ditto.
20471         (rs6000_init_hard_regno_mode_ok): Ditto.
20472         (rs6000_option_override_internal): Ditto.
20473         (legitimate_lo_sum_address_p): Ditto.
20474         (rs6000_legitimize_address): Ditto.
20475         (rs6000_legitimize_reload_address): Ditto.
20476         (rs6000_legitimate_address_p): Ditto.
20477         (abi_v4_pass_in_fpr): Ditto.
20478         (setup_incoming_varargs): Ditto.
20479         (rs6000_gimplify_va_arg): Ditto.
20480         (rs6000_split_multireg_move): Ditto.
20481         (rs6000_savres_strategy): Ditto.
20482         (rs6000_emit_prologue_components): Ditto.
20483         (rs6000_emit_epilogue_components): Ditto.
20484         (rs6000_emit_prologue): Ditto.
20485         (rs6000_emit_epilogue): Ditto.
20486         (rs6000_elf_file_end): Ditto.
20487         (rs6000_function_value): Ditto.
20488         (rs6000_libcall_value): Ditto.
20489         * config/rs6000/rs6000.h: Ditto.
20490         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
20491         (TARGET_MINMAX): ... this.  New.
20492         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
20493         * config/rs6000/rs6000.md: Remove Xilinx FP support.
20494         (*movsi_internal1_single): Delete.
20495         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
20496         mfpu=, mxilinx-fpu): Delete.
20497         * config/rs6000/singlefp.h: Delete.
20498         * config/rs6000/sysv4.h: Remove Xilinx FP support.
20499         * config/rs6000/t-rs6000: Ditto.
20500         * config/rs6000/t-xilinx: Delete.
20501         * config/rs6000/titan.md: Adjust for fp_type removal.
20502         * config/rs6000/vsx.md: Remove Xilinx FP support.
20503         (VStype_simple): Delete.
20504         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
20505         * config/rs6000/xfpu.h: Delete.
20506         * config/rs6000/xfpu.md: Delete.
20507         * config/rs6000/xilinx.h: Delete.
20508         * config/rs6000/xilinx.opt: Delete.
20509         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
20510         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
20512 2018-05-04  Tom de Vries  <tom@codesourcery.com>
20514         PR libgomp/85639
20515         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
20516         if ignore == 0.
20518 2018-05-04  Richard Biener  <rguenther@suse.de>
20520         PR middle-end/85627
20521         * tree-complex.c (update_complex_assignment): We are always in SSA form.
20522         (expand_complex_div_wide): Likewise.
20523         (expand_complex_operations_1): Likewise.
20524         (expand_complex_libcall): Preserve EH info of the original stmt.
20525         (tree_lower_complex): Handle removed blocks.
20526         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
20527         on complex multiplication and division libcall builtins.
20529 2018-05-04  Richard Biener  <rguenther@suse.de>
20531         PR middle-end/85574
20532         * fold-const.c (negate_expr_p): Restrict negation of operand
20533         zero of a division to when we know that can happen without
20534         overflow.
20535         (fold_negate_expr_1): Likewise.
20537 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
20539         PR libstdc++/85466
20540         * real.h (real_nextafter): Declare.
20541         * real.c (real_nextafter): New function.
20542         * fold-const-call.c (fold_const_nextafter): New function.
20543         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
20544         CASE_CFN_NEXTTOWARD.
20545         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
20546         even when arg1_mode is different from arg0_mode.
20548 2018-05-03  Nathan Sidwell  <nathan@acm.org>
20550         * doc/extend.texi (Deprecated Features): Remove
20551         -ffriend-injection.
20552         (Backwards Compatibility): Likewise.
20553         * doc/invoke.texi (C++ Language Options): Likewise.
20554         (C++ Dialect Options): Likewise.
20556 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
20558         PR target/85530
20559         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
20560         _mm512_mask_mullox_epi64): New intrinsics.
20562 2018-05-03  Tom de Vries  <tom@codesourcery.com>
20564         PR testsuite/85106
20565         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20566         dump files): Add offload-tree.
20568 2018-05-03  Richard Biener  <rguenther@suse.de>
20570         PR tree-optimization/85615
20571         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
20572         to loops not nested in BBs loop father to avoid creating multi-entry
20573         loops.
20575 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20577         PR tree-optimization/70291
20578         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
20579         arguments.  Change return type to tree.  Emit libcall as a new
20580         statement rather than replacing existing one when inplace_p is true.
20581         (expand_complex_multiplication_components): New function.
20582         (expand_complex_multiplication): Expand floating-point complex
20583         multiplication using the above.
20584         (expand_complex_division): Rename inner_type parameter to type.
20585         Update expand_complex_libcall call-site.
20586         (expand_complex_operations_1): Update expand_complex_multiplication
20587         and expand_complex_division call-sites.
20589 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
20591         PR target/85582
20592         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
20593         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
20594         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
20595         the highest significant bit of the shift count mask is clear.  In
20596         check whether and[sq]i3 is needed verify that all significant bits
20597         of the shift count other than the highest are set.
20599 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20601         PR libgomp/82428
20602         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
20603         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
20604         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
20605         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
20606         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
20607         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
20608         __builtin_goacc_parlevel_size.
20610 2018-05-02  Richard Biener  <rguenther@suse.de>
20612         PR tree-optimization/85597
20613         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
20614         do not use split vect_get_vec_defs call but call vect_get_slp_defs
20615         directly.
20617 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20619         PR testsuite/85106
20620         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20621         dump files): Add ltrans-tree.
20623 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20625         PR testsuite/85106
20626         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20627         dump files): Add wpa-ipa.
20629 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
20631         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
20632         powerpc*-*-linux*paired* target.
20633         * config/rs6000/750cl.h: Delete.
20634         * config/rs6000/paired.h: Delete.
20635         * config/rs6000/paired.md: Delete.
20636         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
20637         float support.
20638         * config/rs6000/rs6000-builtin.def: Remove paired float support.
20639         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20640         comment.  Remove paired float support.
20641         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
20642         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
20643         VECTOR_PAIRED.
20644         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
20645         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
20646         declarations.
20647         * config/rs6000/rs6000.c: Remove paired float support.
20648         (paired_expand_vector_init, paired_expand_vector_move,
20649         paired_emit_vector_compare, paired_emit_vector_cond_expr,
20650         (paired_expand_lv_builtin, paired_expand_stv_builtin,
20651         paired_expand_builtin, paired_expand_predicate_builtin,
20652         paired_init_builtins): Delete.
20653         * config/rs6000/rs6000.h: Remove paired float support.
20654         * config/rs6000/rs6000.md: Remove paired float support.
20655         (move_from_CR_ov_bit): Delete.
20656         * config/rs6000/rs6000.opt (mpaired): Delete.
20657         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
20658         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
20660 2018-05-02  Richard Biener  <rguenther@suse.de>
20662         PR middle-end/85567
20663         * gimplify.c (gimplify_save_expr): When in SSA form allow
20664         SAVE_EXPRs to compute to SSA vars.
20666 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
20668         PR target/85582
20669         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
20670         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
20671         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
20672         clobber operands[2], instead use a new pseudo.  Formatting fixes.
20674 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
20676         PR tree-optimization/85586
20677         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
20678         exit early for statements in the same group if the accesses are
20679         not strided.
20681 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20683         PR lto/85451
20684         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
20685         error message.
20687 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
20689         PR tree-optimization/85143
20690         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
20692 2018-05-01  Tom de Vries  <tom@codesourcery.com>
20694         PR lto/85451
20695         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
20696         not found" error message.
20698 2018-05-01  Tom de Vries  <tom@codesourcery.com>
20700         PR other/83786
20701         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
20702         * vec.c (test_ordered_remove_if): New function.
20703         (vec_c_tests): Call test_ordered_remove_if.
20704         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
20705         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
20706         * tree-vect-patterns.c (vect_pattern_recog_1): Use
20707         VEC_ORDERED_REMOVE_IF.
20709 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
20711         PR tree-optimization/82665
20712         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
20713         pointer subtraction where arguments come from a memchr call.
20715 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
20717         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
20718         --push-state --as-needed and --pop-state instead of --as-needed and
20719         --no-as-needed if ld supports it.
20720         * configure: Regenerated.
20722         PR web/85578
20723         * doc/install.texi2html: Replace _002d with - and _002a with * in
20724         generated html files using sed.
20726 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20728         PR c++/85523
20729         * gcc-rich-location.c (blank_line_before_p): New function.
20730         (use_new_line): New function.
20731         (gcc_rich_location::add_fixit_insert_formatted): New function.
20732         * gcc-rich-location.h
20733         (gcc_rich_location::add_fixit_insert_formatted): New function.
20735 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20737         * selftest.c (assert_streq): Rename "expected" and "actual" to
20738         "val1" and "val2".  Extend NULL-handling to cover both inputs
20739         symmetrically, while still requiring both to be non-NULL for a pass.
20740         * selftest.h (assert_streq): Rename "expected" and "actual" to
20741         "val1" and "val2".
20742         (ASSERT_EQ): Likewise.
20743         (ASSERT_EQ_AT): Likewise.
20744         (ASSERT_KNOWN_EQ): Likewise.
20745         (ASSERT_KNOWN_EQ_AT): Likewise.
20746         (ASSERT_NE): Likewise.
20747         (ASSERT_MAYBE_NE): Likewise.
20748         (ASSERT_MAYBE_NE_AT): Likewise.
20749         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
20750         the assertion to pass.
20751         (ASSERT_STREQ_AT): Likewise.
20753 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
20755         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
20756         interaction with -pie.
20758 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20760         * selftest.h: Fix alphabetization of per-source-file selftest
20761         declarations.
20763 2018-04-30  Jason Merrill  <jason@redhat.com>
20765         PR c++/61982 - dead stores to destroyed objects.
20766         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
20767         of clobber.
20769 2018-04-30  Jason Merrill  <jason@redhat.com>
20771         * tree.c (build_clobber): New.
20772         * tree.h: Declare it.
20773         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
20775 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20777         * diagnostic-show-locus.c (layout::layout): Update for
20778         location_get_source_line returning a char_span.
20779         (struct char_span): Move to input.h.
20780         (struct correction): Update for fields in char_span becoming
20781         private.
20782         (struct source_line): Update for location_get_source_line
20783         returning a char_span.
20784         (layout::print_line): Likewise.
20785         * edit-context.c (edited_file::print_content): Likewise.
20786         (edited_file::print_diff_hunk): Likewise.
20787         (edited_file::print_run_of_changed_lines): Likewise.
20788         (edited_file::get_num_lines): Likewise.
20789         (edited_line::edited_line): Likewise.
20790         * final.c (asm_show_source): Likewise.
20791         * input.c (location_get_source_line): Convert return type
20792         from const char * to char_span, losing the final "line_len"
20793         param.
20794         (dump_location_info): Update for the above.
20795         (get_substring_ranges_for_loc): Likewise.  Use a char_span
20796         when handling the literal within the line.
20797         (test_reading_source_line): Update for location_get_source_line
20798         returning a char_span.
20799         * input.h (class char_span): Move here from
20800         diagnostic-show-locus.c, converting from a struct to a class.
20801         Make data members private.
20802         (char_span::operator bool): New.
20803         (char_span::length): New.
20804         (char_span::get_buffer): New.
20805         (char_span::operator[]): New.
20806         (char_span::subspan): Make const.
20807         (char_span::xstrdup): New.
20808         (location_get_source_line): Convert return type from const char *
20809         to char_span, losing the final "line_size" param.
20811 2018-04-30  Jan Hubicka  <jh@suse.cz>
20813         * lto-wrapper.c (ltrans_priorities): New static var.
20814         (cmp_priority): New.
20815         (run_gcc): Read priorities and if doing parallel build order
20816         the Makefile by them.
20818 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20820         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
20822 2018-04-30  Richard Biener  <rguenther@suse.de>
20824         * tree-cfg.c (verify_address): Remove base argument, add
20825         flag whether to check TREE_ADDRESSABLE and do that.
20826         (verify_expr): Remove.
20827         (verify_types_in_gimple_reference): Add pieces from verify_expr.
20828         (verify_gimple_assign_single): Likewise.
20829         (verify_gimple_switch): Likewise.
20830         (verify_expr_location_1): Dereference tp once.  Add (disabled)
20831         piece from verify_expr.
20832         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
20834 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20836         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
20838 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20840         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
20841         (small_data_pattern): Likewise.
20842         (arc_rewrite_small_data): Likewise.
20843         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
20844         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
20845         (get_symbol_alignment): New function.
20846         (legitimate_small_data_address_p): Likewise.
20847         (legitimate_scaled_address): Update, call
20848         legitimate_small_data_address_p.
20849         (output_sdata): New static variable.
20850         (arc_print_operand): Update how we handle small data operands.
20851         (arc_print_operand_address): Likewise.
20852         (arc_legitimate_address_p): Update, use
20853         legitimate_small_data_address_p.
20854         (arc_rewrite_small_data_p): Remove.
20855         (arc_rewrite_small_data_1): Likewise.
20856         (arc_rewrite_small_data): Likewise.
20857         (small_data_pattern): Likewise.
20858         (compact_sda_memory_operand): Update to use
20859         legitimate_small_data_address_p and get_symbol_alignment.
20860         (prepare_move_operands): Don't rewite sdata pattern.
20861         (prepare_extend_operands): Remove.
20862         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
20863         pattern.
20864         (zero_extendqisi2): Likewise.
20865         (zero_extendhisi2): Likewise.
20866         (extendqihi2): Likewise.
20867         (extendqisi2): Likewise.
20868         (extendhisi2): Likewise.
20869         (addsi3): Likewise.
20870         (subsi3): Likewise.
20871         (andsi3): Likewise.
20872         * config/arc/constraints.md (Usd): Change it to memory constraint.
20874 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20876         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
20877         as source of std instructions.
20878         * config/arc/arc.md (movsi_insn): Update pattern predicate to
20879         allow 6-bit constants as source for store instructions.
20880         (movdi_insn): Update instruction pattern to allow 6-bit constants
20881         as source for store instructions.
20883 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
20885         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
20887 2018-04-30  Nathan Sidwell  <nathan@acm.org>
20888             Sandra Loosemore <sandra@codesourcery.com>
20890         * dumpfile.c (dump_open): Allow '-' for stdout.
20891         * doc/invoke.texi (Developer Options): Document dump filename
20892         determination early.  Document stdin/stdout selection.
20894 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
20896         Microblaze Target: PIC data text relative
20898         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
20899         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
20900         Add declaration.
20901         * config/microblaze/microblaze.h (microblaze_constant_address_p):
20902         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
20903         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
20904         New addressing mode for data-text relative position indepenedent code.
20905         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
20906         'ADDRESS_SYMBOLIC_TXT_REL'.
20907         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
20908         (microblaze_legitimate_pic_operand): Exclude function calls from
20909         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
20910         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
20911         addresses cases.
20912         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
20913         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
20914         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
20915         for 'address + offset'.
20916         (microblaze_expand_prologue): Add new function prologue call for
20917         'r20' assignation.
20918         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
20919         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
20920         table in case of TARGET_PIC_DATA_TEXT_REL.
20921         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
20922         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
20923         Add new macros 'UNSPEC_TEXT',
20924         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
20925         + exclude function calls from 'UNSPEC_PLT' in case of data text
20926         relative mode.
20927         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
20928         new target hook for generating address diff vector tables in case of
20929         flag_pic.
20930         * doc/tm.texi : Regenerate.
20931         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
20932         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
20933         of addr diff vector generation.
20934         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
20935         target hook definition.
20936         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
20937         Add default function for generate_pic_addr_diff_vec -> flag_pic.
20938         * doc/invoke.texi (Add new pic option): Add new microblaze pic
20939         option for data text relative.
20941 2018-04-30  Richard Biener  <rguenther@suse.de>
20943         * tree-chrec.h (evolution_function_is_constant_p): Remove
20944         redundant check.
20945         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
20947 2018-04-30  Richard Biener  <rguenther@suse.de>
20949         PR bootstrap/85571
20950         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
20952 2018-04-30  Richard Biener  <rguenther@suse.de>
20954         PR tree-optimization/28364
20955         PR tree-optimization/85275
20956         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
20957         copying first exit test.
20959 2018-04-28  Mark Wielaard  <mark@klomp.org>
20961         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
20962         dwarf_version >= 5.
20963         (dwarf_AT): Handle DW_AT_addr_base.
20964         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
20966 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
20968         PR target/84431
20969         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
20970         (*ashl<dwi>3_doubleword_mask_1): Ditto.
20971         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
20972         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
20974 2018-04-28  Richard Biener  <rguenther@suse.de>
20976         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
20977         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
20978         to reflect use.  Only add interesting stmts.
20980 2018-04-27  Martin Jambor  <mjambor@suse.cz>
20982         PR ipa/85549
20983         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
20984         the jump function allows for passing through aggregate values.
20986 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
20988         * input.h (in_system_header_at): Convert from macro to inline
20989         function.
20990         (from_macro_expansion_at): Likewise.
20991         (from_macro_definition_at): Likewise.
20993 2018-04-27  Jeff Law  <law@redhat.com>
20995         * config.gcc: Mark tile* targets as deprecated/obsolete.
20997 2018-04-27  Richard Biener  <rguenther@suse.de>
20999         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
21000         fix for ILP32.
21002 2018-04-27  Richard Biener  <rguenther@suse.de>
21004         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
21006 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
21008         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
21009         with Yd constraint. Set "preferred_for_speed" attribute from
21010         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
21011         with Yd constraint.
21012         (*movdi_internal): Ditto.
21013         (movti_interunit splitters): Remove
21014         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
21015         (movdi_interunit splitters): Ditto.
21016         * config/i386/constraints.md (Ye): Remove.
21017         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
21019 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21021         PR target/85512
21022         * config/aarch64/constraints.md (Usg): Limit to 31.
21023         (Usj): Limit to 63.
21025 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
21027         PR tree-optimization/85529
21028         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
21029         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
21030         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
21031         zero extension or masking of the MSB bit.
21032         (optimize_range_tests): Add FIRST_BB argument, pass it through
21033         to optimize_range_tests_var_bound.
21034         (maybe_optimize_range_tests, reassociate_bb): Adjust
21035         optimize_range_tests callers.
21037 2018-04-26  Richard Biener  <rguenther@suse.de>
21038             Jakub Jelinek  <jakub@redhat.com>
21040         * cgraph.h (symbol_table): Just declare debug method here.
21041         * symtab.c (symbol_table::debug): Define.
21043 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
21045         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
21047 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
21049         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
21050         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
21051         (*movdi_internal): Substitute Yi and Yj constraint with x
21052         and Ym and Yn constraint with y constraint.  Update "isa"
21053         attribute and set "preferred_for_speed" attribute from
21054         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
21055         (*movsi_internal): Ditto.
21056         (*movdf_internal): Ditto.
21057         (*movsf_internal): Ditto.
21058         (*zero_extendsidi2): Ditto.
21059         * config/i386/sse.md (vec_set<mode>_0): Ditto.
21060         (sse2_loadld): Ditto.
21061         (*vec_extract<ssevecmodelower>_0): Ditto.
21062         (*vec_extractv4si_0_zext_sse4): Ditto.
21063         (vec_concatv2di): Ditto.
21064         (*vec_dup<mode>): Ditto.
21065         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
21066         * config/i386/constraints.md (Yi): Remove.
21067         (Yj): Remove.
21068         (Ym): Remove.
21069         (Yn): Remove.
21071 2018-04-26  Nathan Sidwell  <nathan@acm.org>
21073         * dumpfile.c (dump_open): New.
21074         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
21075         (dump_finish): Detect stdio/stderr by value not name.
21077 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
21079         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
21081 2018-04-26  Tom de Vries  <tom@codesourcery.com>
21083         PR target/84952
21084         * config/nvptx/nvptx.c (verify_neutering_jumps)
21085         (verify_neutering_labels): New function
21086         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
21088 2018-04-26  Tom de Vries  <tom@codesourcery.com>
21090         PR target/84025
21091         * config/nvptx/nvptx.c (needs_neutering_p): New function.
21092         (nvptx_single): Use needs_neutering_p to skip over insns that do not
21093         need neutering.
21095 2018-04-26  Richard Biener <rguenther@suse.de>
21096             Tom de Vries  <tom@codesourcery.com>
21098         PR lto/85422
21099         * lto-streamer-out.c (output_function): Fixup loops if required to match
21100         discovery done in the reader.
21102 2018-04-26  Richard Biener  <rguenther@suse.de>
21104         PR tree-optimization/85116
21105         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
21106         have a loop exit from the single latch predecessor.  Remove
21107         case of header with just condition.
21108         (ch_base::copy_headers): Exclude infinite loops from any
21109         processing.
21110         (pass_ch::execute): Record exits.
21112 2018-04-26  Richard Biener  <rguenther@suse.de>
21114         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
21115         prologue cost vector and pass it to vect_get_load_cost.
21116         (vect_get_peeling_costs_all_drs): Likewise.
21117         (vect_peeling_hash_get_lowest_cost): Likewise.
21118         (vect_enhance_data_refs_alignment): Likewise.
21120 2018-04-26  Richard Biener  <rguenther@suse.de>
21122         PR middle-end/85450
21123         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
21124         checking of integer<->pointer conversions.
21125         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
21126         sign-/zero-extending pointer types.
21127         (expand_omp_for_static_chunk): Likewise.
21129 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
21130             Jean Lee  <xiaoyur347@gmail.com>
21132         * config/mips/mips.c (mips_asan_shadow_offset): New function.
21133         (TARGET_ASAN_SHADOW_OFFSET): Define.
21134         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
21135         true for -fsanitize=address.
21137 2018-04-25  Mark Wielaard  <mark@klomp.org>
21139         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
21140         shorter ones.
21142 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
21144         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
21145         than "alu", remove explicit "memory" and "imm_disp" attributes.
21146         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
21148         PR middle-end/85414
21149         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
21150         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
21151         gen_lowpart_no_emit.
21153 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
21155         PR target/85473
21156         * config/i386/i386.c (ix86_expand_builtin): Change memory
21157         operand to XI, extend p0 to Pmode.
21158         * config/i386/i386.md: Change unspec volatile and operand
21159         1 mode to XI, change operand 0 mode to P.
21161 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21163         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
21164         GET_MODE_MASK before any checking.
21165         (nds32_can_use_bset_p): Likewise.
21166         (nds32_can_use_btgl_p): Likewise.
21168 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21170         * config/nds32/nds32-doubleword.md: New define_split pattern for
21171         illegal register number.
21173 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21175         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
21177 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21179         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
21181 2018-04-25  Richard Biener  <rguenther@suse.de>
21183         * lto-streamer.h (LTO_major_version): Bump to 8.
21185 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
21187         * BASE-VER: Set to 9.0.0.
21189 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
21191         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
21192         in __abskf2 and __powikf2.
21194 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21196         PR target/85512
21197         * config/aarch64/constraints.md (Usg, Usj): New constraints.
21198         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
21199         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
21200         Use the above on operand 2.  Reindent.
21201         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
21203 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
21205         PR target/85485
21206         * common/config/i386/i386-common.c (ix86_handle_option): Don't
21207         handle OPT_mcet.
21208         * config/i386/i386.opt (mcet): Removed.
21209         * doc/install.texi: Remove -mcet documentation.
21210         * doc/invoke.texi: Likewise.
21212 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
21214         PR target/85485
21215         * doc/install.texi: Remove -mcet from bootstrap-cet.
21217 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
21219         PR target/85511
21220         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
21221         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
21222         if TARGET_64BIT.
21224         PR target/85503
21225         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
21226         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
21227         containing a CONST_VECTOR.
21229 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
21231         * doc/install.texi: Update newlib dependency for nvptx.
21233 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
21235         PR target/85508
21236         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
21237         instead of INTVAL when shifting x left.
21239 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
21241         PR tree-optimization/85478
21242         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
21243         vect_grouped_store_supported for single element vectors.
21245 2018-04-24  Richard Biener  <rguenther@suse.de>
21247         PR target/85491
21248         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
21249         load cost increase to the case of non-constant step.
21251 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
21253         PR target/84828
21254         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
21255         destination if any_malformed_asm.
21257 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
21259         PR middle-end/85496
21260         * expr.c (store_field): In the bitfield case, if the value comes from
21261         a function call and is returned in registers by means of a PARALLEL,
21262         do not change the mode of the temporary unless BLKmode and VOIDmode.
21264 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
21266         PR rtl-optimization/85423
21267         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
21268         dependencies to debug insns when the previous insn is non-debug.
21270 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
21272         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
21273         enums into a single definition.
21274         (fls): Fix predicates and printing.
21275         (seti): Likewise.
21277 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
21279         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
21280         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
21281         and short u6 immediate.
21282         (check_if_valid_sleep_operand): Remove.
21283         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
21285 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21287         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
21288         flag_always_save_lp condition.
21289         * config/nds32/nds32.opt (malways-save-lp): New option.
21291 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
21293         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
21294         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
21295         * config/nds32/nds32.h
21296         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
21297         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
21299 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
21301         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
21302         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
21304 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
21305             Chung-Ju Wu  <jasonwucj@gmail.com>
21307         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
21308         Declare.
21309         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
21310         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
21312 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21314         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
21316 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21318         * config/nds32/nds32-protos.h (nds32_data_alignment,
21319         nds32_local_alignment): Declare.
21320         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
21321         nds32_local_alignment): New functions.
21322         (TARGET_CONSTANT_ALIGNMENT): Define.
21323         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
21325 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21327         * config/nds32/nds32.c
21328         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
21329         (TARGET_MODES_TIEABLE_P): Likewise.
21331 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21333         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
21334         level Ofast and Og.
21336 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
21337             Chung-Ju Wu  <jasonwucj@gmail.com>
21339         * config/nds32/constants.md (unspec_volatile_element): Add enum values
21340         for unaligned access.
21341         * config/nds32/nds32-intrinsic.c: Implementation of expanding
21342         unaligned access.
21343         * config/nds32/nds32-intrinsic.md: Likewise.
21344         * config/nds32/nds32_intrinsic.h: Likewise.
21345         * config/nds32/nds32.h (nds32_builtins): Likewise.
21346         * config/nds32/nds32.opt (munaligned-access): New option.
21347         * config/nds32/nds32.c (nds32_asm_file_start): Display
21348         flag_unaligned_access status.
21350 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
21352         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
21353         -mno-relax is present.
21354         * config/riscv/linux.h (LINK_SPEC): Ditto.
21356 2018-04-20  Martin Sebor  <msebor@redhat.com>
21358         PR c/85365
21359         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
21360         for null pointers.
21361         (gimple_fold_builtin_stxcpy_chk): Same.
21362         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
21364 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
21366         PR target/85456
21367         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
21368         __powikf2 when long double is IEEE 128-bit.
21370 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
21372         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
21373         step to make sure stack always aligned.
21375 2018-04-20  Carl Love  <cel@us.ibm.com>
21377         PR target/83402
21378         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
21379         size check for arg0.
21381 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
21382             Tom de Vries  <tom@codesourcery.com>
21384         PR target/85445
21385         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
21386         Emit insns for calls too.
21387         (nvptx_find_par): Always look for worker-level predecessor insn.
21388         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
21389         calls.
21390         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
21391         (nvptx_process_pars): Propagate frames for calls.
21393 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
21395         PR target/85469
21396         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
21397         Removed.
21398         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
21399         (ix86_handle_option): Don't handle OPT_mibt.
21400         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
21401         __SHSTK__.
21402         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
21403         has_ibt and ibt.
21404         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
21405         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
21406         (ix86_target_macros): Define __CET__ with flag_cf_protection
21407         for -fcf-protection.
21408         * config/i386/i386.c (isa2_opts): Remove -mibt.
21409         * config/i386/i386.h (TARGET_IBT): Removed.
21410         (TARGET_IBT_P): Likewise.
21411         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
21412         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
21413         * config/i386/i386.opt (mcet): Update help message.
21414         (mshstk): Likewise.
21415         (mibt): Removed.
21416         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
21417         -mcet as an alias for -mshstk.
21419 2018-04-20  Richard Biener <rguenther@suse.de>
21421         PR middle-end/85475
21422         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
21423         complexity by forcing a single use of the multiply operand.
21425 2018-04-20  Martin Jambor  <mjambor@suse.cz>
21427         ipa/85449
21428         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
21429         recursion dependency to only apply to non-clones.
21431 2018-04-20  Martin Jambor  <mjambor@suse.cz>
21433         ipa/85447
21434         * ipa-cp.c (create_specialized_node): Check that clones of
21435         self-recursive edges exist during IPA-CP.
21437 2018-04-19  Toon Moene  <toon@moene.org>
21439         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
21440         by -O3.
21442 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
21444         PR tree-optimization/85467
21445         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
21446         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
21447         VECTOR_CST element to type.
21449 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
21451         PR target/85397
21452         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
21453         * config/i386/i386.md (builtin_setjmp_setup): Removed.
21454         (builtin_longjmp): Likewise.
21455         (save_stack_nonlocal): New pattern.
21456         (restore_stack_nonlocal): Likewise.
21458 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
21460         PR target/85404
21461         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
21462         Replace ASM_OUTPUT_LABEL with fprintf.
21464 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
21466         PR target/85417
21467         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
21468         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
21469         * config/i386/i386-c.c (ix86_target_macros_internal): Also
21470         define __IBT__ and __SHSTK__ for -fcf-protection.
21471         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
21472         TARGET_IBT.
21473         (ix86_trampoline_init): Likewise.
21474         (x86_output_mi_thunk): Likewise.
21475         (ix86_notrack_prefixed_insn_p): Likewise.
21476         (ix86_option_override_internal): Don't disallow -fcf-protection.
21477         * config/i386/i386.md (rdssp<mode>): Also enable for
21478         -fcf-protection.
21479         (incssp<mode>): Likewise.
21480         (nop_endbr): Likewise.
21481         * config/i386/i386.opt (mcet): Change help message to built-in
21482         functions only.
21483         (mibt): Likewise.
21484         (mshstk): Likewise.
21485         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
21486         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
21487         enable CET built-in functions.
21489 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
21491         * common/config/i386/i386-common.c
21492         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
21493         OPTION_MASK_ISA_MOVDIRI_UNSET,
21494         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
21495         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
21496         * config.gcc (movdirintrin.h): New header.
21497         * config/i386/cpuid.h (bit_MOVDIRI,
21498         bit_MOVDIR64B): New bits.
21499         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
21500         and -mmvodir64b.
21501         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
21502         (VOID, PVOID, PCVOID)): New function types.
21503         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
21504         __builtin_ia32_directstoreu_u64,
21505         __builtin_ia32_movdir64b): New builtins.
21506         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
21507         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
21508         and -mmovdiri.
21509         (ix86_valid_target_attribute_inner_p): Ditto.
21510         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
21511         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
21512         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
21513         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
21514         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
21515         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
21516         (movdiri<mode>, movdir64b_<mode>): New.
21517         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
21518         * config/i386/immintrin.h: Include movdirintrin.h.
21519         * config/i386/movdirintrin.h: New file.
21520         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
21522 2018-04-19  Richard Biener  <rguenther@suse.de>
21524         PR middle-end/85455
21525         * cfg.c (clear_bb_flags): When loop state says we have
21526         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
21528 2018-04-19  Richard Biener  <rguenther@suse.de>
21530         PR tree-optimization/84737
21531         * tree-vect-data-refs.c (vect_copy_ref_info): New function
21532         copying restrict info.
21533         (vect_setup_realignment): Use it.
21534         * tree-vectorizer.h (vect_copy_ref_info): Declare.
21535         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
21536         the first DR to all generated stores.
21537         (vectorizable_load): Likewise for loads.
21539 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
21541         PR tree-optimization/85446
21542         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
21543         the integral and pointer types to have the same precision.
21545         * doc/install.texi: Document --disable-cet being the default and
21546         --enable-cet=auto.
21548 2018-04-18  Martin Liska  <mliska@suse.cz>
21550         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
21551         style.
21553 2018-04-18  Martin Liska  <mliska@suse.cz>
21555         Revert
21556         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
21558         PR ipa/83983
21559         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
21560         arguments if they are comparable.
21562 2018-04-18  Martin Liska  <mliska@suse.cz>
21564         Revert
21565         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
21567         PR lto/84805
21568         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
21569         incomplete types.
21571 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
21573         PR target/85388
21574         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
21575         ENDBR after calling __morestack.
21577 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
21579         PR jit/85384
21580         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
21581         by using gcc_base_ver to generate a gcc_driver_version, and use
21582         it when generating GCC_DRIVER_NAME.
21583         * configure: Regenerate.
21585 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
21587         PR target/81084
21588         * config.gcc: Obsolete powerpc*-*-*spe*.
21590 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21592         PR debug/84637
21593         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
21594         (stabstr_D): Change type of unum from unsigned int to
21595         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
21596         type.
21598 2018-04-17  Jim Wilson  <jimw@sifive.com>
21600         PR 84856
21601         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
21602         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
21603         Set arg_pointer_offset after using pretend_args_size.
21605 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21607         PR rtl-optimization/85431
21608         * dse.c (record_store): Ignore zero width stores.
21610         PR sanitizer/85230
21611         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
21612         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
21613         __builtin_stack_restore rather than after it.
21614         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
21615         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
21616         argument instead of virtual_dynamic_stack_rtx.
21618 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21620         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
21621         New prototype.
21622         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21623         Add note to error message to explain internal mapping of overloaded
21624         built-in function name to non-overloaded built-in function name.
21625         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
21626         function.
21628 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
21630         PR target/85424
21631         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
21632         where the inputs overlap with the output.
21634 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21636         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
21637         (=v, v) alternative and explicit "memory" attribute.
21638         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
21639         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21640         attributes.
21641         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
21642         "sselog1" type instead of "sselog".
21643         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
21644         "sselog".  Remove explicit "memory" attribute.
21645         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
21646         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21647         attributes.
21648         (vec_extract_hi_v32hi): Merge all alternatives into one, use
21649         "sselog1" type instead of "sselog".  Remove explicit "memory"
21650         attribute.
21651         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
21652         use "sselog1" type instead of "sselog".  Remove explicit "memory"
21653         attribute.
21654         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
21655         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21656         attributes.
21657         (vec_extract_hi_v64qi): Merge all alternatives into one, use
21658         "sselog1" type instead of "sselog".  Remove explicit "memory"
21659         attribute.
21660         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
21661         use "sselog1" type instead of "sselog".  Remove explicit "memory"
21662         attribute.
21664         PR target/85430
21665         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
21667         PR middle-end/85414
21668         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
21669         on a SUBREG.
21671 2018-04-17  Martin Jambor  <mjambor@suse.cz>
21673         PR ipa/85421
21674         * ipa-cp.c (create_specialized_node): Call
21675         expand_all_artificial_thunks if necessary.
21677 2018-04-17  Martin Liska  <mliska@suse.cz>
21679         PR lto/85405
21680         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
21681         in message, remote space in between '_G' and '('.
21683 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21685         PR target/85281
21686         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
21687         avx512f_vmcmp<mode>3<round_saeonly_name>,
21688         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
21689         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
21690         avx512f_rndscale<mode><round_saeonly_name>,
21691         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
21692         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
21693         Use %<iptr>2 instead of %2 for -masm=intel.
21694         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
21695         avx512f_vcvttss2usi<round_saeonly_name>,
21696         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
21697         -masm=intel.
21698         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
21699         avx512f_vcvttsd2usi<round_saeonly_name>,
21700         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
21701         Use %q1 instead of %1 for -masm=intel.
21702         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
21703         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
21704         of %3 for -masm=intel.
21705         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
21706         -masm=intel.
21707         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
21708         -masm=intel.
21709         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
21710         -masm=intel.
21711         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
21712         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
21713         %g1.
21714         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
21715         -masm=intel.
21716         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
21717         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
21718         %g1 and one with %0 and %1.
21719         (avx512er_vmrcp28<mode><round_saeonly_name>,
21720         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
21721         %1 for -masm=intel.
21722         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
21723         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
21724         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
21725         of %0 and %{%4%} for -masm=intel.
21726         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
21727         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
21728         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
21729         order of %0 and %{%5%}%{z%} for -masm=intel.
21731 2018-04-17  Jan Hubicka  <jh@suse.cz>
21733         PR lto/85405
21734         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
21736 2018-04-17  Martin Liska  <mliska@suse.cz>
21738         PR ipa/85329
21739         * multiple_target.c (create_dispatcher_calls): Set apostrophes
21740         for target_clone error message.  Make default implementation
21741         clone to be a local declaration.
21742         (separate_attrs): Add new argument and check for an empty
21743         string.
21744         (expand_target_clones): Handle it.
21745         (ipa_target_clone): Make redirection just for target_clones
21746         functions.
21748 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
21749             Tom de Vries  <tom@codesourcery.com>
21751         PR middle-end/84955
21752         * omp-expand.c (expand_oacc_for): Add dummy false branch for
21753         tiled basic blocks without omp continue statements.
21755 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
21757         PR target/83660
21758         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
21759         vec_extract expression as having side effects to make sure it gets
21760         a cleanup point.
21762 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
21764         PR target/85403
21765         * config/i386/i386.c (get_builtin_code_for_version): Check
21766         error_mark_node.
21768 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
21770         PR target/84331
21771         * config.gcc: Support "skylake".
21772         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
21773         PROCESSOR_SKYLAKE.
21774         * config/i386/i386.c (m_SKYLAKE): Define.
21775         (processor_target_table): Add "skylake".
21776         (ix86_option_override_internal): Add "skylake".
21777         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
21778         PROCESSOR_CANNONLAKE.
21779         (get_builtin_code_for_version): Fix priority for
21780         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
21781         PROCESSOR_SKYLAKE-AVX512.
21782         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
21783         (processor_type): Add PROCESSOR_SKYLAKE.
21785 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
21786             Jason Merrill  <jason@redhat.com>
21788         PR c++/85112
21789         * convert.c (convert_to_integer_1): Use direct recursion for
21790         enumeral types and types with a precision less than the number
21791         of bits in their mode.
21793 2018-04-16  Julia Koval  <julia.koval@intel.com>
21795         PR target/84413
21796         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
21797         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
21799 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
21801         PR target/85293
21802         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
21803         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
21804         and -mno-direct-move.
21806 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
21808         PR target/83402
21809         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
21810         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
21811         Ensure negative shifts result in {0}.
21813 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
21815         PR rtl-optimization/79916
21816         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
21817         regs (if any) to define how to gnerate SD moves when LRA is in
21818         progress.
21820 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
21822         PR rtl-optimization/85393
21823         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
21824         * except.c (expand_dw2_landing_pad_for_region): Make static.
21825         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
21826         a label and unconditional jump to old_bb, rather than
21827         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
21828         basic block.
21830         PR rtl-optimization/85376
21831         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
21832         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
21833         instead of a specific value.
21835 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
21836             Bin Cheng  <bin.cheng@arm.com>
21838         PR tree-optimization/82965
21839         PR tree-optimization/83991
21840         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
21841         by_profile_only parameter.
21842         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
21843         information if the loop was predicted to iterate too many times.
21844         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
21846 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
21848         PR lto/71991
21849         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
21850         always inline.
21852 2018-04-13  Martin Liska  <mliska@suse.cz>
21853             Jakub Jelinek  <jakub@redhat.com>
21855         PR middle-end/81657
21856         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
21857         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
21858         * builtins.c (expand_builtin_memory_copy_args): Use
21859         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
21860         handle dest_addr == pc_rtx.
21862 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
21864         PR target/85291
21865         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
21866         asked to not generate direct moves.
21867         (fix_trunc<mode>si2_stfiwx): Similar.
21868         (fix_trunc<mode>si2_internal): Similar.
21870 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21872         PR debug/83157
21873         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
21874         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
21875         lookup if dest in some wider mode is known to be const0_rtx and
21876         if so, record permanent equivalence for it to be ZERO_EXTEND of
21877         the narrower mode destination.
21879 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
21881         * lto-streamer-out.c (output_function): Revert 259346.
21882         * omp-expand.c (expand_oacc_for): Likewise.
21884 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
21886         PR rtl-optimization/85354
21887         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
21888         * sel-sched.c (sel_global_init): ... here.
21890 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
21892         PR target/85238
21893         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
21894         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
21895         mode for PE-COFF targets.
21896         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
21897         (i386_pe_asm_lto_end): Likewise.
21898         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
21899         (TARGET_ASM_LTO_END): Likewise.
21900         * config/i386/winnt.c (saved_debug_info_level): New static variable.
21901         (i386_pe_asm_lto_start): New function.
21902         (i386_pe_asm_lto_end): Likewise.
21904 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
21905             Richard Biener  <rguenther@suse.de>
21907         PR middle-end/84955
21908         * lto-streamer-out.c (output_function): Fix CFG loop state before
21909         streaming out.
21910         * omp-expand.c (expand_oacc_for): Handle calls to internal
21911         functions like regular functions.
21913 2018-04-12  Richard Biener  <rguenther@suse.de>
21915         PR lto/85371
21916         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
21917         for the early LTO debug to properly generate references to it
21918         during DIE emission.  Do not re-use that for the skeleton for
21919         split-dwarf.
21920         (dwarf2out_early_finish): Likewise.
21922 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21924         PR target/85328
21925         * config/i386/sse.md
21926         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
21927         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
21928         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
21929         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
21930         and output is a reg, avoid creating invalid lowpart subreg, but
21931         instead split into a 512-bit move.  Don't split if not AVX512VL,
21932         input is xmm16+ reg and output is a mem.
21933         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
21934         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
21935         xmm16+ reg and output is a mem.
21937 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
21939         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
21940         also for flag_dwarf2_cfi_asm.
21942 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21944         PR rtl-optimization/85342
21945         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
21946         a bool scalar var inside of the loop instead.  Don't try to update
21947         recog_data.operand after failed apply_change_group.
21949 2018-04-12  Tom de Vries  <tom@codesourcery.com>
21951         PR target/85296
21952         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
21953         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
21954         array with flexible array member as array without given dimension.
21955         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
21956         argument for undefined param to true.
21958 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
21960         PR target/85321
21961         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
21962         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
21963         from PowerPC section.
21964         * config/rs6000/sysv4.opt (mcall-): Improve help text.
21965         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
21966         help text that is too long.
21967         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
21968         help text that is too long.
21969         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
21970         help text that is too long.
21972 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
21974         * config/alpha/alpha.md (stack_probe_internal): Rename
21975         from "probe_stack".  Update all callers.
21977 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
21979         PR rtl-optimization/84566
21980         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
21981         sched_macro_fuse_insns.
21983 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
21985         PR target/84301
21986         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
21987         (compute_block_dependences): ... from here.
21989 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
21991         PR tree-optimization/85331
21992         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
21993         from int to HOST_WIDE_INT.
21995 2018-04-11  Martin Jambor  <mjambor@suse.cz>
21997         PR ipa/84149
21998         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
21999         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
22000         not the same as the source val.
22001         (cgraph_edge_brings_value_p): New parameter.
22002         (gather_edges_for_value): Pass destination value to
22003         cgraph_edge_brings_value_p.
22004         (perhaps_add_new_callers): Likewise.
22005         (get_info_about_necessary_edges): Likewise and exclude values brought
22006         only by self-recursive edges.
22007         (create_specialized_node): Redirect only clones of self-calling edges.
22008         (+self_recursive_pass_through_p): New function.
22009         (find_more_scalar_values_for_callers_subset): Use it.
22010         (find_aggregate_values_for_callers_subset): Likewise.
22011         (known_aggs_to_agg_replacement_list): Removed.
22012         (decide_whether_version_node): Re-calculate known constants for all
22013         remaining context clones.
22015 2018-04-11  Richard Biener  <rguenther@suse.de>
22017         PR lto/85339
22018         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
22019         from early DWARF output.
22020         (dwarf2out_early_finish): Output line info unconditionally into
22021         early DWARF and add reference to it.
22023 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
22025         PR target/85281
22026         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
22027         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
22028         other than V2DFmode using iptr mode attribute.
22029         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
22031 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
22033         PR rtl-optimization/84659
22034         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
22036 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
22038         PR debug/85302
22039         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
22040         SIZEP is NULL.
22041         (output_loc_list): Pass address of a dummy size variable even in the
22042         locview handling loop.
22043         (index_location_lists): Add comment on why skip_loc_list_entry can't
22044         call size_of_locs.
22046 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
22048         PR target/85261
22049         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
22050         into register.
22052 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
22054         PR target/85321
22055         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
22056         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
22057         and -mstring-compare-inline-limit.
22059 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
22061         PR target/85287
22062         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
22063         for stack clash protection in a register whenever we need it to be in
22064         a register.
22066 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
22068         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
22069         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
22071 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
22073         PR target/85321
22074         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
22075         the help text.
22076         (mlong-double-): Ditto.
22077         * config/rs6000/sysv4.opt (msdata=): Ditto.
22078         (mtls-size=): Ditto.
22080 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22082         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
22083         erroneous entries for
22084         "vector int vec_ldl (int, long int *)", and
22085         "vector unsigned int vec_ldl (int, unsigned long int *)".
22086         Add comments and entries for
22087         "vector bool char vec_ldl (int, bool char *)",
22088         "vector bool short vec_ldl (int, bool short *)",
22089         "vector bool int vec_ldl (int, bool int *)",
22090         "vector bool long long vec_ldl (int, bool long long *)",
22091         "vector pixel vec_ldl (int, pixel *)",
22092         "vector long long vec_ldl (int, long long *)",
22093         "vector unsigned long long vec_ldl (int, unsigned long long *)".
22094         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
22095         type tree bool_long_long_type_node and correct definition of
22096         bool_V2DI_type_node to make reference to this new type tree.
22097         (rs6000_mangle_type): Replace erroneous reference to
22098         bool_long_type_node with bool_long_long_type_node.
22099         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
22100         comments to emphasize sign distinctions for char and int types and
22101         replace RS6000_BTI_bool_long constant with
22102         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
22103         use of RS6000_BTI_pixel.
22104         (bool_long_type_node): Remove this macro definition.
22105         (bool_long_long_type_node): New macro definition
22107 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
22109         PR rtl-optimization/85300
22110         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
22111         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
22112         simplify_unary_operation fails.
22114 2018-04-10  Martin Liska  <mliska@suse.cz>
22116         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
22117         cgraph_edge and ipa_ref.
22119 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
22121         PR target/85177
22122         PR target/85255
22123         * config/i386/sse.md
22124         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
22125         computation of the VEC_MERGE selector from mask.
22126         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
22127         Fix decoding of the VEC_MERGE selector into mask.
22129 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
22131         PR tree-optimization/85286
22132         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
22134 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
22136         * final.c (final_1): Set insn_last_address as well as
22137         insn_current_address.
22139 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22141         PR target/85173
22142         * explow.c (emit_stack_probe): Call validize_mem on memory location
22143         before passing it to gen_probe_stack.  Create address operand and
22144         legitimize it for the probe_stack_address case.
22146 2018-04-09  Jan Hubicka  <jh@suse.cz>
22148         PR lto/85078
22149         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
22150         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
22151         * tree.c (free_lang_data_in_type): Fix handling of binfos;
22152         walk basetypes.
22153         (free_lang_data): Rebuild type inheritance graph.
22155 2018-04-09  Martin Sebor  <msebor@redhat.com>
22157         * invoke.texi (-finline-small-functions): Mention other optimization
22158         options.
22159         (-findirect-inlining, -fpartial-inlining): Same.
22160         (-finline-functions-called-once): Same.
22161         (-freorder-blocks-and-partition): Same.
22163 2018-04-09  Jan Hubicka  <jh@suse.cz>
22165         PR rtl/84058
22166         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
22167         jumps; choose last target that matches the criteria (i.e.
22168         no partition changes for non-crossing jumps).
22169         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
22170         support for redirecting crossing jumps to non-crossing.
22172 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
22174         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
22175         also for naked functions.
22177 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
22179         * config/arc/arc.md (add_shift): New pattern.
22180         (add_shift2): Likewise.
22181         (sub_shift): Likewise.
22182         (sub_shift_cmp0_noout): Likewise.
22183         (compare_si_ashiftsi): Likewise.
22184         (xbfu_cmp0_noout): New combine pattern.
22185         (xbfu_cmp0"): Likewise.
22186         (movsi_set_cc_insn): Place the predicable variant first.
22187         (commutative_binary_cmp0_noout): Remove clobber.
22188         (commutative_binary_cmp0): New pattern.
22189         (noncommutative_binary_cmp0): Likewise.
22190         (noncommutative_binary_cmp0_noout): Likewise.
22191         (noncommutative_binary_comparison_result_used): Removed.
22192         (rsub_cmp0): New pattern.
22193         (rsub_cmp0_noout): Likewise.
22194         (extzvsi): Changed, keep only meaningful variants.
22195         (SQH, SEZ): New iterators.
22196         (SQH_postfix): New mode attribute.
22197         (SEZ_prefix): New code attribute.
22198         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
22199         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
22200         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
22201         of numerical value.
22202         (noncommutative_operator): Check the availability of barrel
22203         shifter option.
22205 2018-04-09  Richard Biener  <rguenther@suse.de>
22207         PR tree-optimization/85284
22208         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
22209         Only use the niter constraining form of simple_iv when the exit
22210         is always executed.
22212 2018-04-09  Tom de Vries  <tom@codesourcery.com>
22214         PR target/84041
22215         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
22216         (define_expand "*memory_barrier"): New define_expand.
22217         (define_insn "memory_barrier"): New insn.
22219 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22221         PR rtl-optimization/80463
22222         PR rtl-optimization/83972
22223         PR rtl-optimization/83480
22225         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
22226         correct producer for the insn.
22227         (tidy_control_flow): Fixup seqnos in case of debug insns.
22229 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22231         PR rtl-optimization/83913
22233         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
22234         different sched-times when merging exprs.
22236 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22238         PR rtl-optimization/83962
22240         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
22241         tidy_fallthru_edge and tidy_control_flow.
22243 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22245         PR rtl-optimization/83530
22247         * sel-sched.c (force_next_insn): New global variable.
22248         (remove_insn_for_debug): When force_next_insn is true, also leave only
22249         next insn in the ready list.
22250         (sel_sched_region): When the region wasn't scheduled, make another pass
22251         over it with force_next_insn set to 1.
22253 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
22255         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
22256         into tm_file.
22257         * config/nds32/constants.md (unspec_volatile_element): Add enum values
22258         for interrupt control.
22259         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
22260         functions for interrupt control.
22261         * config/nds32/nds32-intrinsic.md: Likewise.
22262         * config/nds32/nds32_intrinsic.h: Likewise.
22263         * config/nds32/nds32.h (nds32_builtins): Likewise.
22265 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
22267         * config/nds32/nds32.c (nds32_init_machine_status,
22268         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
22269         strict_aligned_p field.
22270         (nds32_expand_to_rtl_hook): New function.
22271         (TARGET_EXPAND_TO_RTL_HOOK): Define.
22272         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
22274 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22275             Chung-Ju Wu  <jasonwucj@gmail.com>
22277         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
22278         * config/nds32/nds32-n7.md: New file.
22279         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
22280         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
22281         pipeline.
22282         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
22283         * config/nds32/nds32.md (pipeline_model): Add n7.
22284         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
22285         * config/nds32/pipelines.md: Include n7 settings.
22287 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22288             Chung-Ju Wu  <jasonwucj@gmail.com>
22290         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
22291         * config/nds32/nds32-e8.md: New file.
22292         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
22293         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
22294         pipeline.
22295         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
22296         * config/nds32/nds32.md (pipeline_model): Add e8.
22297         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
22298         * config/nds32/pipelines.md: Include e8 settings.
22300 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22301             Chung-Ju Wu  <jasonwucj@gmail.com>
22303         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
22304         * config/nds32/nds32-n8.md: New file.
22305         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
22306         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
22307         pipeline.
22308         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
22309         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
22310         * config/nds32/nds32.md (pipeline_model): Add n8.
22311         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
22312         * config/nds32/pipelines.md: Include n8 settings.
22314 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22315             Chung-Ju Wu  <jasonwucj@gmail.com>
22317         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
22318         * config/nds32/nds32-n9-2r1w.md: New file.
22319         * config/nds32/nds32-n9-3r2w.md: New file.
22320         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
22321         nds32_register_ports): New or modify for cpu n9.
22322         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
22323         pipeline.
22324         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
22325         * config/nds32/nds32-utils.c: New file.
22326         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
22327         TARGET_MUL_SLOW): Define.
22328         * config/nds32/nds32.md (pipeline_model): New attribute.
22329         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
22330         New options that support cpu n9.
22331         * config/nds32/pipelines.md: Include n9 settings.
22332         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
22334 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
22336         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
22337         information if necessary.
22338         (output_cond_branch_compare_zero): Likewise.
22339         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
22340         (nds32_target_alignment): Refine for alignment.
22341         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
22342         (FUNCTION_BOUNDARY): Modify.
22343         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
22344         align case.
22345         * config/nds32/nds32.opt (malways-align, malign-functions): New.
22347 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
22349         * config/nds32/constants.md (unspec_volatile_element): Add values for
22350         TLB operation and data prefetch.
22351         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
22352         functions for TLB operation and data prefetch.
22353         * config/nds32/nds32-intrinsic.md: Likewise.
22354         * config/nds32/nds32_intrinsic.h: Likewise.
22355         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
22356         (nds32_print_operand): Likewise.
22357         * config/nds32/nds32.h (nds32_builtins): Likewise.
22359 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
22360         Andrew Pinski <pinsika@gcc.gnu.org>
22362         PR middle-end/82976
22363         * match.pd: Use constant_boolean_node of correct type instead of
22364         boolean_true_node or boolean_false_node for simplifying
22365         pointer comparisons to zero.
22367 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
22369         PR tree-optimization/80021
22370         * tree.c (verify_type_variant): Make error call in verify_variant_match
22371         translatable and remove final full stop.
22373 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22375         * config/nds32/constants.md (unspec_volatile_element): Add
22376         UNSPEC_VOLATILE_EH_RETURN.
22377         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
22378         nds32_output_stack_pop): Support dwarf exception handling process.
22379         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
22380         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
22381         exception handling process.
22382         (nds32_compute_stack_frame): Likewise.
22383         (nds32_return_addr_rtx): Likewise.
22384         (nds32_initial_elimination_offset): Likewise.
22385         (nds32_expand_prologue): Likewise.
22386         (nds32_expand_epilogue): Likewise.
22387         (nds32_dynamic_chain_address): New function.
22388         * config/nds32/nds32.h (machine_function): Add fields for dwarf
22389         exception handling.
22390         (DYNAMIC_CHAIN_ADDRESS): Define.
22391         (EH_RETURN_DATA_REGNO): Define.
22392         (EH_RETURN_STACKADJ_RTX): Define.
22393         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
22394         patterns for dwarf exception handling.
22396 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22398         * config/nds32/nds32.h: Clean up obsolete macros.
22400 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22402         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
22403         Add enum values for particular instructions.
22404         * config/nds32/nds32-intrinsic.c: Implementation of expanding
22405         particular intrinsic functions.
22406         * config/nds32/nds32-intrinsic.md: Likewise.
22407         * config/nds32/nds32_intrinsic.h: Likewise.
22408         * config/nds32/nds32.h (nds32_builtins): Likewise.
22409         * config/nds32/nds32.md (type): Add pbsad and pbsada.
22410         (btst, ave): New patterns for particular instructions.
22412 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22414         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
22415         Add enum values for atomic load/store and memory sync.
22416         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
22417         and memory sync.
22418         * config/nds32/nds32-intrinsic.md: Likewise.
22419         * config/nds32/nds32_intrinsic.h: Likewise.
22420         * config/nds32/nds32.h (nds32_builtins): Likewise.
22422 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
22424         PR tree-optimization/85257
22425         * fold-const.c (native_encode_vector): If not all elts could fit
22426         and off is -1, return 0 rather than offset.
22427         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
22428         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
22429         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
22430         adjust buffer in native_interpret_expr call.
22432 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22434         * config/nds32/constants.md (unspec_volatile_element): Add cache
22435         control enum values.
22436         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
22437         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
22438         * config/nds32/nds32.c (nds32_cctl_names): New.
22439         (nds32_print_operand): Handle cache control register names.
22440         * config/nds32/nds32.h (nds32_builtins): New enum values.
22441         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
22442         macros.
22443         * config/nds32/nds32.md (type): Add mmu.
22444         * config/nds32/pipelines.md (simple_insn): Add mmu.
22446 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22448         * config/nds32/nds32.md (type): Remove call.
22449         * config/nds32/pipelines.md (simple_insn): Likewise.
22451 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22453         * config/nds32/constants.md (unspec_volatile_element): Add
22454         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
22455         UNSPEC_VOLATILE_FMFCFG.
22456         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
22457         description for fmfcfg and fmfcsr.
22458         (bdesc_1arg): Add fmtcsr.
22459         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
22460         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
22461         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
22462         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
22463         unspec_fmfcfg): New patterns.
22464         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
22465         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
22466         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
22467         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
22468         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
22469         __nds32__fmfcfg): Define.
22471 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22473         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
22474         intrinsic register names.
22475         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
22476         intrinsic register enum values and macros.
22478 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22480         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
22481         for load/store addressing form.
22482         (nds32_print_operand_address): Likewise.
22484 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
22486         PR target/85196
22487         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
22488         based on LABEL_REF.  Remove useless assertion.
22489         (pic_address_needs_scratch): Fix formatting.
22490         (sparc_legitimize_pic_address): Minor tweaks.
22491         (sparc_delegitimize_address): Adjust assertion accordingly.
22492         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
22493         into symbolic_operand.
22494         (movsi_high_pic_label_ref): Likewise.
22495         (movsi_lo_sum_pic_label_ref): Likewise.
22496         (movdi_pic_label_ref): Likewise.
22497         (movdi_high_pic_label_ref): Likewise.
22498         (movdi_lo_sum_pic_label_ref): Likewise.
22500 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
22502         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
22503         custom LIB_SPEC setup.
22505 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
22506             Kito Cheng  <kito.cheng@gmail.com>
22508         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
22509         * config/riscv/freebsd.h: New.
22511 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22513         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
22514         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
22515         file.
22517 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22518             Kito Cheng  <kito.cheng@gmail.com>
22520         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
22521         nds32_output_call, nds32_symbol_binds_local_p): New functions.
22522         * config/nds32/nds32-protos.h (nds32_output_call,
22523         nds32_output_return): Declare.
22524         * config/nds32/nds32.md: Refine all the call and return patterns.
22526 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
22528         PR debug/85252
22529         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
22530         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
22532         PR rtl-optimization/84872
22533         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
22534         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
22535         EDGE_CROSSING edge.
22537 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
22539         * expr.c (copy_blkmode_to_reg): Revert 254862.
22540         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
22542 2018-04-06  Richard Biener  <rguenther@suse.de>
22544         PR middle-end/85244
22545         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
22546         after seeing a component reference with an adjacent field.  Treat
22547         refs to arrays at struct end of external decls similar to
22548         refs to unconstrained commons.
22550 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
22552         PR sanitizer/85213
22553         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
22554         look through SAVE_EXPRs with non-side-effects argument.  Adjust
22555         recursive calls.
22556         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
22557         save_p here.
22559 2018-04-06  Richard Biener  <rguenther@suse.de>
22561         PR middle-end/85180
22562         * alias.c (find_base_term): New wrapper around find_base_term
22563         unwinding CSELIB_VAL_PTR changes.
22564         (find_base_term): Do not restore CSELIB_VAL_PTR during the
22565         recursion.
22567 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22569         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
22570         instructions.
22571         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
22572         constant definitions.
22573         ("nop"): lr 0,0 -> nopr r0
22574         ("nop_lr0", "nop_lr1"): New insn definitions.
22576 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22578         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
22579         NDS32_V3PUSH_AVAILABLE_P macro.
22581 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
22582             Chung-Ju Wu  <jasonwucj@gmail.com>
22584         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
22585         (nds32*-*-*): Add float and fpu_config into supported_defaults.
22586         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
22587         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
22588         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
22589         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
22590         * config/nds32/constraints.md: New constraints and checking for hard
22591         float configuration.
22592         * config/nds32/iterators.md: New mode iterator and attribute for hard
22593         float configuration.
22594         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
22595         patterns.
22596         * config/nds32/nds32-fpu.md: New file.
22597         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
22598         deal with hard float code generation.
22599         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
22600         ARCH_V3S.
22601         (abi_type, float_reg_number): New enum type.
22602         * config/nds32/nds32-predicates.c: New predicates for hard float.
22603         * config/nds32/nds32-protos.h: Declare functions for hard float.
22604         * config/nds32/nds32.c: Implementation for hard float configuration.
22605         * config/nds32/nds32.h: Definitions for hard float configuration.
22606         * config/nds32/nds32.md: Include hard float machine description and
22607         modify patterns for hard float configuration.
22608         * config/nds32/nds32.opt: New options for hard float configuration.
22609         * config/nds32/predicates.md: New predicates for hard float
22610         configuration.
22612 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
22614         * common/config/nds32/nds32-common.c
22615         (nds32_option_optimization_table): Enable -mreleax-hint by default.
22617 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
22619         PR middle-end/85195
22620         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
22621         CONSTRUCTOR_ELT (ctor, ...)->value.
22623 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
22625         PR target/85193
22626         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
22628 2018-04-05  Tom de Vries  <tom@codesourcery.com>
22630         PR target/85204
22631         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
22632         cond jump.
22634 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
22635             Kito Cheng  <kito.cheng@gmail.com>
22637         * config/nds32/constraints.md (U33): Fine-tune checking condition.
22638         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
22639         * config/nds32/nds32.h (nds32_16bit_address_type): Add
22640         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
22642 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
22643             Kito Cheng  <kito.cheng@gmail.com>
22645         * config/nds32/constraints.md (Ufe): New memory constraint.
22646         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
22647         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
22648         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
22649         operands.
22650         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
22651         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
22653 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22655         * config/nds32/nds32.md: Use optimize_size in the condition for
22656         alu-shift instructions.
22658 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22660         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
22662 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22664         * config/nds32/nds32.md (negsi2): Refine pattern.
22666 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
22667             Chung-Ju Wu  <jasonwucj@gmail.com>
22669         * config/nds32/iterators.md (shift_rotate): New code iterator.
22670         (shift): New code attribute.
22671         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
22672         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
22673         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
22674         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
22675         bit-wise operations.
22676         (andsi3, *andsi3): Ditto.
22677         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
22678         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
22679         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
22680         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
22681         nds32_ior_operand, nds32_xor_operand): New predicates.
22683 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22685         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
22686         (addsi3, subsi3): ... this.
22688 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22690         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
22692 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22694         * config/nds32/nds32.md: Adjust indention.
22696 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
22698         * config/nds32/nds32.md (feature): New attribute.
22700 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22702         * config/nds32/nds32.md (subtype): New attribute.
22704 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
22706         PR target/85203
22707         * config/arm/arm-builtins.c (arm_expand_builtin): Change
22708         expansion to perform a bitwise AND of the argument followed by a
22709         boolean negation of the result.
22711 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
22713         PR rtl-optimization/84878
22714         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
22715         the basic block.  Assert the use reference is not artificial and that
22716         it has an associated insn.
22718 2018-04-04  Michael Matz  <matz@suse.de>
22720         * builtins.c (compute_objsize): Pass correct operand
22721         to array_at_struct_end_p.
22723 2018-04-04  Richard Biener  <rguenther@suse.de>
22725         PR lto/85176
22726         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
22727         from contexts for DINFO_LEVEL_TERSE and below.
22729 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22731         * config/nds32/nds32-doubleword.md (move_<mode>): Require
22732         resiter_operand condition.
22733         * config/nds32/nds32.md (*move<mode>): Ditto.
22735 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22736             Monk Chiang  <sh.chiang04@gmail.com>
22738         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
22740 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22742         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
22744 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22745             Kito Cheng  <kito.cheng@gmail.com>
22747         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
22748         nds32_cond_code_str, output_cond_branch,
22749         output_cond_branch_compare_zero, nds32_expand_cbranch,
22750         nds32_expand_cstore, nds32_expand_movcc,
22751         nds32_output_cbranchsi4_equality_zero,
22752         nds32_output_cbranchsi4_equality_reg,
22753         nds32_output_cbranchsi4_equality_reg_or_const_int,
22754         nds32_output_cbranchsi4_greater_less_zero: New functions.
22755         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
22756         nds32_expand_cstore, nds32_expand_movcc,
22757         nds32_output_cbranchsi4_equality_zero,
22758         nds32_output_cbranchsi4_equality_reg,
22759         nds32_output_cbranchsi4_equality_reg_or_const_int,
22760         nds32_output_cbranchsi4_greater_less_zero): Declare.
22761         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
22762         nds32_rimm11s_operand): New predicates.
22763         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
22764         * config/nds32/nds32.md: Rewrite all the branch and conditional move
22765         patterns.
22767 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22769         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
22770         * config/nds32/nds32.md: Ditto.
22771         * config/nds32/pipelines.md: Ditto.
22773 2018-04-04  Richard Biener  <rguenther@suse.de>
22775         PR tree-optimization/85168
22776         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
22777         propagating abnormals.
22779 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22781         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
22783 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22784             Kito Cheng  <kito.cheng@gmail.com>
22786         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
22787         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
22788         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
22789         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
22790         * config/nds32/nds32.md (sibcall_internal): New.
22791         (sibcall_register): Remove.
22792         (sibcall_immediate): Remove.
22793         (sibcall_value_internal): New.
22794         (sibcall_value_register): Remove.
22795         (sibcall_value_immediate): Remove.
22796         * config/nds32/predicates.md (nds32_general_register_operand): New.
22797         (nds32_call_address_operand): New.
22799 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
22801         PR rtl-optimization/85167
22802         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
22803         bb_defs if *split_p, instead preinitialize it to NULL.
22805         PR tree-optimization/85156
22806         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
22807         evaluating the argument multiple times.
22809 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
22811         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
22812         than vector.
22813         (_mm_cvtpd_ps): Likewise.
22814         (_mm_cvttpd_epi32): Likewise.
22815         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
22816         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
22817         vector, pixel, and bool following altivec.h include.
22819 2018-04-03  Martin Sebor  <msebor@redhat.com>
22821         * doc/extend.texi (Common Function Attributes): Clarify.
22822         (const attribute): Likewise.
22823         (pure attribute): Likewise.
22825 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
22827         PR target/85169
22828         * config/i386/i386.c (ix86_expand_vector_set): Use
22829         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
22831 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
22833         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
22834         instructions when changing rounding bits to preserve precision bits
22835         in the x87 control word.
22837 2018-04-03  Martin Liska  <mliska@suse.cz>
22839         PR tree-optimization/82491
22840         * rtl.h (strip_offset_and_add): Replace += suboffset with
22841         poly_uint64 () + suboffset.
22843 2018-03-29  Martin Liska  <mliska@suse.cz>
22844             Martin Jambor  <mjambor@suse.cz>
22846         PR ipa/84947
22847         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
22848         param_type is not an integral or pointer type.
22850 2018-04-03  Richard Biener  <rguenther@suse.de>
22852         * sese.h (recompute_all_dominators): Remove.
22854 2018-04-02  Martin Sebor  <msebor@redhat.com>
22856         * doc/invoke.texi (-Wrestrict): Fix typos.
22858 2018-04-02  Jim Wilson  <jimw@sifive.com>
22860         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
22861         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
22862         (<optab>di3, <optab>si3_extend): Likewise.
22863         (<optab>si3_mask, <optab>si3_mask_1): New.
22864         (<optab>di3_mask, <optab>di3_mask_1): New.
22865         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
22866         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
22867         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
22869 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
22871         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
22872         example.
22874 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
22876         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
22877         (nds32_canonicalize_comparison): New function.
22879 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22880             Kito Cheng  <kito.cheng@gmail.com>
22881             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
22883         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
22884         * config/nds32/constants.md (unspec_volatile_element): Add
22885         UNSPEC_VOLATILE_RELAX_GROUP.
22886         * config/nds32/nds32-relax-opt.c: New file.
22887         * config/nds32/nds32-predicates.c
22888         (nds32_symbol_load_store_p): New function.
22889         * config/nds32/nds32-protos.h
22890         (nds32_symbol_load_store_p): Declare function.
22891         (make_pass_nds32_relax_opt): Declare new rtl pass function.
22892         * config/nds32/nds32.c
22893         (nds32_register_pass): New function to register pass.
22894         (nds32_register_passes): New function to register passes.
22895         * config/nds32/nds32.md (relax_group): New pattern.
22896         * config/nds32/nds32.opt (mrelax-hint): New option.
22897         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
22899 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
22901         * config/nds32/t-nds32: Modify files dependency.
22903 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22905         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
22906         (PROFILE_HOOK): Define its implementation.
22908 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22910         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
22911         type and 32-bit size.
22913 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
22915         PR middle-end/85090
22916         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
22917         (V_128_256): New mode iterator.
22918         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
22919         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
22920         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
22921         of V.
22922         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
22923         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
22925 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
22927         PR target/83315
22928         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
22929         NaN inputs correctly.
22931 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
22933         PR target/80546
22934         * config/rs6000/vsx.md (??r): New mode attribute.
22935         (*vsx_mov<mode>_64bit): Use it.
22936         (*vsx_mov<mode>_32bit): Likewise.
22938 2018-03-30  Martin Sebor  <msebor@redhat.com>
22940         PR tree-optimization/84818
22941         * builtins.c (check_access): Use warning_n.
22943 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
22945         PR target/83822
22946         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
22947         condition.
22948         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
22949         condition.
22951 2018-03-30  Julia Koval  <julia.koval@intel.com>
22953         PR target/84413
22954         * x86-tune.def (movx, partial_reg_dependency): Enable for
22955         m_SKYLAKE_AVX512.
22957 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
22959         PR inline-asm/84985
22960         * lra-constraints.c (process_alt_operands): Move setting
22961         this_alternative_matches below.
22963 2018-03-29  Martin Liska  <mliska@suse.cz>
22965         PR lto/84995.
22966         * doc/invoke.texi: Document how LTO works with debug info.
22967         Describe auto-load support of binutils.  Mention 'x86-64'
22968         as valid option value of -march option.
22970 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
22972         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
22974         PR c/85094
22975         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
22976         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
22977         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
22978         checking.
22980 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
22982         PR target/84912
22983         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
22984         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
22985         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
22986         for RS6000_BTM_POWERPC64.
22987         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
22988         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
22989         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
22990         definition.
22991         (DIVDE): Use it.
22992         (DIVDEU): Likewise.
22994 2018-03-28  Carl Love  <cel@us.ibm.com>
22996         Revert
22997         2017-09-27  Carl Love  <cel@us.ibm.com>
22999         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
23000         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
23001         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
23002         fctiw instruction.
23004 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23006         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
23007         instead of __vector bool.
23008         (_mm_max_pu8): Likewise.
23009         (_mm_min_pi16): Likewise.
23011 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
23013         PR target/84912
23014         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
23015         (DIVWEUO): Likewise.
23016         (DIVDEO): Likewise.
23017         (DIVDEUO): Likewise.
23018         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
23019         DIVWEUO and DIVDEUO.
23020         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
23021         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
23022         (div_extend): Likewise.
23023         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
23024         builtin function.
23025         (__builtin_divweuo): Likewise.
23026         (__builtin_divdeo): Likewise.
23027         (__builtin_divdeuo): Likewise.
23029 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
23031         PR target/85095
23032         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
23033         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
23035         PR tree-optimization/82004
23036         * gimple-match-head.c (optimize_pow_to_exp): New function.
23037         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
23038         Don't fold to exp if optimize_pow_to_exp is false.
23040 2018-03-28  Martin Liska  <mliska@suse.cz>
23042         PR other/84819
23043         * calls.c (initialize_argument_information): Fix trailing space.
23044         * common.opt: Fix typo and provide better explanation for
23045         -fsanitize-coverage option.
23046         * config/i386/i386.opt: Fix typo.
23048 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
23049             Martin Liska  <mliska@suse.cz>
23051         PR sanitizer/85081
23052         * gimplify.c (asan_poison_variable): Don't do the check for
23053         gimplify_omp_ctxp here.
23054         (gimplify_decl_expr): Do it here.
23055         (gimplify_target_expr): Likewise.
23057 2018-03-28  Martin Liska  <mliska@suse.cz>
23059         PR target/84988
23060         * config/i386/i386.c (ix86_function_arg_advance): Do not call
23061         chkp_type_bounds_count if MPX is not enabled.
23063 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
23065         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
23067 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
23069         PR target/84914
23070         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
23071         function to create the function decl for complex long double
23072         multiply and divide for -mabi=ieeelongdouble.
23073         (init_float128_ieee): Call it.
23075 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
23077         PR target/85044
23078         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
23079         -fcf-protection=branch -mibt.
23080         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
23082 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
23084         PR target/81863
23085         * config/arm/arm.c (arm_valid_symbolic_address): Handle
23086         arm_word_relocations.
23088 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
23090         PR target/85056
23091         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
23092         extern array declarations.
23094 2018-03-27  Richard Biener  <rguenther@suse.de>
23096         PR middle-end/84067
23097         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
23098         explicit single_use checks.
23100 2018-03-27  Richard Biener  <rguenther@suse.de>
23102         PR tree-optimization/85082
23103         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
23104         Valueize the VUSE.
23106 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
23108         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
23109         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
23110         Turn on fasynchronous-unwind-tables and funwind-tables.
23112 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
23114         PR target/85073
23115         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
23116         (*bmi_blsr_<mode>_ccz): Ditto.
23118 2018-03-26  Tom de Vries  <tom@codesourcery.com>
23120         PR tree-optimization/85063
23121         * omp-general.c (offloading_function_p): New function.  Factor out
23122         of ...
23123         * omp-offload.c (pass_omp_target_link::gate): ... here.
23124         * omp-general.h (offloading_function_p): Declare.
23125         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
23126         with attribute omp declare target for offloading functions.
23128 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
23130         PR tree-optimization/84005
23131         * tree-data-ref.h (get_base_for_alignment): Declare.
23132         * tree-data-ref.c (get_base_for_alignment_1): New function.
23133         (get_base_for_alignment): Likewise.
23134         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
23135         get_base_for_alignment to find a suitable base object, instead
23136         of always using drb->base_address.
23138 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
23140         PR inline-asm/85022
23141         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
23142         known size by default.
23144 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
23146         PR inline-asm/85030
23147         * lra-constraints.c (process_alt_operands): Don't match BLKmode
23148         and non BLKmode operands.
23150 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23152         PR target/85026
23153         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
23154         Clean up attributes.
23156 2018-03-23  Richard Biener  <rguenther@suse.de>
23158         PR debug/85020
23159         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
23160         we are going to emit early debug for LTO.
23162 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
23164         PR inline-asm/85034
23165         * function.c (match_asm_constraints_1): Don't optimize if input
23166         doesn't satisfy general_operand predicate for output's mode.
23168         PR inline-asm/85022
23169         * alias.c (write_dependence_p): Don't require for x_canonicalized
23170         non-VOIDmode if x has VOIDmode.
23172         PR sanitizer/85029
23173         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
23174         just don't try to optimize it rather than assert it never happens.
23176 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23178         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
23179         macro expansions for definition of ST_INTERNAL_<mode> and
23180         LD_INTERNAL_<mode> builtins.
23181         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
23182         Remove prototype.
23183         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
23184         function.
23185         (altivec_expand_st_builtin): Likewise.
23186         (altivec_expand_builtin): Remove calls to deleted functions.
23187         (rs6000_address_for_altivec): Delete this function.
23188         * config/rs6000/vector.md: Remove expands for
23189         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
23191 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
23193         PR target/84826
23194         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
23195         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
23196         re-computing once computed.
23197         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
23198         (arm_init_machine_status): Initialize
23199         machine->static_chain_stack_bytes.
23201 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23203         PR target/84760
23204         * doc/extend.texi: Add four new prototypes for vec_ld.
23205         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
23206         definitions for more logical presentation.
23207         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
23208         entries for V1TI variants of __builtin_altivec_ld builtin.
23209         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
23210         handling of V1TI variant of LVX icode pattern.
23211         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
23212         (rs6000_gimple_fold_builtin): Likewise.
23213         (altivec_init_builtins): Add code to define
23214         __builtin_altivec_lvx_v1ti function.
23216 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
23218         PR inline-asm/84941
23219         * function.c (match_asm_constraints_1): Don't do the optimization
23220         if input isn't a REG, SUBREG, MEM or constant.
23222 2018-03-22  Tom de Vries  <tom@codesourcery.com>
23224         PR tree-optimization/84956
23225         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
23226         bb_has_abnormal_pred.
23228 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
23230         PR sanitizer/85018
23231         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
23232         DECL_INITIAL (decl) to decl at the end.
23233         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
23234         adjust the comment.
23236 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
23238         * doc/extend.texi (__builtin_tgmath): Document when complex
23239         integer types are treated as _Complex _Float64.
23241 2018-03-21  Tom de Vries  <tom@codesourcery.com>
23243         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
23245 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
23247         PR tree-optimization/84960
23248         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
23249         if it is ENTRY block, move them into single succ of ENTRY in that case.
23251 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
23253         PR tree-optimization/84811
23254         * poly-int.h (poly_span_traits): Remove the T3 parameter and
23255         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
23256         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
23257         (known_subrange_p): Update accordingly.  Cast each value involved
23258         in the size comparison, rather than casting the result of the
23259         subtraction.
23261 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
23263         PR tree-optimization/84982
23264         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
23265         by flipping the least significant bit rather than all bits from
23266         bitpos to bitpos + bitsize - 1.
23268 2018-03-21  Nathan Sidwell  <nathan@acm.org>
23270         * doc/extend.texi (Deprecated Features): Remove mention of
23271         long-deleted deprecations.
23273 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23275         PR jit/84288
23276         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
23277         * configure: Regenerate.
23279 2018-03-21  Tom de Vries  <tom@codesourcery.com>
23281         PR tree-optimization/83126
23282         * tree-parloops.c (num_phis): New function.
23283         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
23285 2018-03-21  Nathan Sidwell  <nathan@acm.org>
23287         * doc/extend.texi (Deprecated Features): Update deprecated flags,
23288         mention anon-struct/union members and trailing attributes.
23290 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
23292         PR tree-optimization/84969
23293         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
23294         builtin memset partitions if they set different rhs values.
23296 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
23298         PR rtl-optimization/84989
23299         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
23300         VEC_DUPLICATE with scalar result mode.
23302 2018-03-21  Martin Liska  <mliska@suse.cz>
23304         PR ipa/84963
23305         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
23306         not intended return statement.
23308 2018-03-21  Martin Liska  <mliska@suse.cz>
23310         PR target/84988
23311         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
23312         (chkp_find_bound_slots_1): Limit number of iterations.
23314 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
23316         PR target/84838
23317         * Minor grammar fixes for x86 options.
23319 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
23321         PR debug/84875
23322         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
23323         holding REG_CFA_RESTORE notes, instead turn them into a USE.
23325 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
23327         PR target/83789
23328         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
23329         (altivec_lvx_<mode>_1op): Likewise.
23330         (altivec_stvx_<mode>_2op): Likewise.
23331         (altivec_stvx_<mode>_1op): Likewise.
23332         (altivec_lvx_<VM2:mode>): New define_expand.
23333         (altivec_stvx_<VM2:mode>): Likewise.
23334         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
23335         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
23336         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
23337         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
23338         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
23339         (rs6000_gen_lvx): Likewise.
23340         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
23341         (altivec_expand_stv_builtin): Likewise.
23342         (altivec_expand_builtin): Likewise.
23343         * config/rs6000/vector.md: Likewise.
23345 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23347         PR target/82518
23348         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
23349         BYTES_BIG_ENDIAN.
23351 2018-03-20  Richard Biener  <rguenther@suse.de>
23353         PR target/84986
23354         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
23355         sign-conversions as zero, fall back to standard scalar_stmt
23356         cost for the rest.
23358 2018-03-20  Martin Liska  <mliska@suse.cz>
23360         PR ipa/84825
23361         * predict.c (rebuild_frequencies): Handle case when we have
23362         PROFILE_ABSENT, but flag_guess_branch_prob is false.
23364 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
23366         PR target/84990
23367         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
23368         flag_section_anchors.
23369         * varasm.c (use_blocks_for_decl_p): Remove hack for
23370         dw2_force_const_mem.
23372         PR target/84845
23373         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
23374         to ...
23375         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
23376         be created, use lowpart_subreg of operands[0] rather than operands[0]
23377         itself.
23378         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
23379         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
23380         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
23381         and n constraint instead of aarch64_shift_imm_di and Usd.
23382         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
23383         (*aarch64_<optab>_reg_minus<mode>3): ... this.
23385 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
23387         PR target/82989
23388         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
23389         to favor GPR over NEON registers.
23390         (<shift>di3_neon): Likewise.
23392 2018-03-20  Tom de Vries  <tom@codesourcery.com>
23394         PR target/84952
23395         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
23396         (nvptx_process_pars): Emit bar.sync asap and alap.
23398 2018-03-20  Tom de Vries  <tom@codesourcery.com>
23400         PR target/84954
23401         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
23402         seen_label if seen_label is already set.
23404 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
23406         PR target/84945
23407         * config/i386/i386.c (fold_builtin_cpu): For features above 31
23408         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
23409         Use 1U instead of 1.  Formatting fixes.
23411         PR c/84953
23412         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
23413         instead of TREE_TYPE (s1) for the return value.
23415 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
23417         PR tree-optimization/84946
23418         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
23419         bitsize + bitsize in poly_uint64 rather than poly_int64.
23421         PR sanitizer/78651
23422         * dwarf2asm.c: Include fold-const.c.
23423         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
23424         of decl rather than decl itself.
23426         PR rtl-optimization/84643
23427         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
23429 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
23431         PR sanitizer/78651
23432         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
23433         calling assemble_variable.
23435 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
23437         PR target/81647
23438         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
23439         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
23441 2018-03-19  Jim Wilson  <jimw@sifive.com>
23443         PR bootstrap/84856
23444         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
23445         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
23446         (riscv_first_stack_step): Likewise.
23447         (riscv_option_override): Use STACK_BOUNDARY instead of
23448         MIN_STACK_BOUNDARY.
23449         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
23450         MIN_STACK_BOUNDARY.
23451         (BIGGEST_ALIGNMENT): Set to 128.
23452         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
23453         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
23454         STACK_BOUNDARY.
23456 2018-03-19  Richard Biener  <rguenther@suse.de>
23458         PR tree-optimization/84933
23459         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
23460         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
23462 2018-03-19  Richard Biener  <rguenther@suse.de>
23464         PR tree-optimization/84859
23465         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
23466         (cond_if_else_store_replacement): Perform sinking operation on
23467         single-store BBs regardless of MAX_STORES_TO_SINK setting.
23468         Generalize what a BB with a single eligible store is.
23470 2018-03-19  Richard Biener  <rguenther@suse.de>
23472         PR tree-optimization/84929
23473         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
23474         chrec_is_positive against non-chrec arg.
23476 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
23478         PR target/84711
23479         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
23481 2018-03-18  Martin Liska  <mliska@suse.cz>
23483         PR rtl-optimization/84635
23484         * regrename.c (build_def_use): Use matches_mode only when
23485         matches >= 0.
23487 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
23489         PR tree-optimization/84913
23490         * tree-vect-loop.c (vectorizable_reduction): Don't try to
23491         vectorize chains of COND_EXPRs.
23493 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23495         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
23497 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23499         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
23501 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23503         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
23505 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
23506             Kito Cheng  <kito.cheng@gmail.com>
23508         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
23509         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
23510         (nds32_adjust_reg_alloc_order): New function.
23511         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
23513 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
23515         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
23516         nds32_print_operand, nds32_print_operand_address): Use
23517         HOST_WIDE_INT_PRINT_DEC instead.
23519 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
23521         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
23523 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
23525         PR target/84902
23526         * config/i386/i386.c (initial_ix86_tune_features,
23527         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
23528         unsigned long long.
23529         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
23530         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
23531         rather than 1u << ix86_tune.  Formatting fix.
23532         (ix86_option_override_internal): Change ix86_arch_mask from
23533         unsigned int to unsigned HOST_WIDE_INT, initialize to
23534         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
23535         (ix86_function_specific_restore): Likewise.
23537 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23539         PR target/84899
23540         * postreload.c (reload_combine_recognize_pattern): Perform
23541         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
23542         truncate_int_for_mode the result for the destination's mode.
23544         PR c/84909
23545         * hsa-gen.c (mem_type_for_type): Fix comment typo.
23546         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
23547         Likewise.
23548         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
23549         Likewise.
23551 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
23553         PR target/84876
23554         * lra-assigns.c (lra_split_hard_reg_for): Don't use
23555         regno_allocno_class_array and sorted_pseudos.
23556         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
23557         insns where regno is used.
23559 2018-03-16  Martin Liska  <mliska@suse.cz>
23561         PR ipa/84833
23562         * multiple_target.c (create_dispatcher_calls): Redirect
23563         reference in the symbol table.
23565 2018-03-16  Martin Liska  <mliska@suse.cz>
23567         PR ipa/84722
23568         * multiple_target.c (create_dispatcher_calls): Redirect also
23569         an alias.
23571 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23573         PR c++/79937
23574         PR c++/82410
23575         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
23576         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
23577         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
23579 2018-03-16  Julia Koval  <julia.koval@intel.com>
23581         * doc/invoke.texi (Skylake Server): Add CLWB.
23582         Cannonlake): Remove CLWB.
23584 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23586         PR tree-optimization/84841
23587         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
23588         1 << 3.
23589         (FLOAT_ONE_CONST_TYPE): Define.
23590         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
23591         (sort_by_operand_rank): Put entries with higher constant_type last
23592         rather than first to match comments.
23594 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
23596         * config/nios2/nios2.md (movsi_internal): Fix thinko in
23597         split predicate.
23599 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
23601         PR c++/79085
23602         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
23603         check and use address of target always.
23605 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
23607         PR target/84574
23608         * config/i386/i386.c (indirect_thunk_needed): Update comments.
23609         (indirect_thunk_bnd_needed): Likewise.
23610         (indirect_thunks_used): Likewise.
23611         (indirect_thunks_bnd_used): Likewise.
23612         (indirect_return_needed): New.
23613         (indirect_return_bnd_needed): Likewise.
23614         (output_indirect_thunk_function): Add a bool argument for
23615         function return.
23616         (output_indirect_thunk_function): Don't generate alias for
23617         function return thunk.
23618         (ix86_code_end): Call output_indirect_thunk_function to generate
23619         function return thunks.
23620         (ix86_output_function_return): Set indirect_return_bnd_needed
23621         and indirect_return_needed instead of indirect_thunk_bnd_needed
23622         and indirect_thunk_needed.
23624 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
23626         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
23627         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
23628         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
23630 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
23631             Paul Hua <paul.hua.gm@gmail.com>
23633         PR c/84852
23634         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
23636 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
23638         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
23639         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
23640         resp. SFmode cases.
23642 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
23644         PR target/84711
23645         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
23646         instead of GET_MODE_SIZE when comparing Units.
23648 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
23650         PR target/68256
23651         * varasm.c (hash_section): Return an unchangeble hash value
23652         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
23653         Return !aarch64_can_use_per_function_literal_pools_p ().
23655 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
23657         PR target/84860
23658         * optabs.c (emit_conditional_move): Pass address of cmode's copy
23659         rather than address of cmode as last argument to prepare_cmp_insn.
23661 2018-03-15  Julia Koval  <julia.koval@intel.com>
23663         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
23664         F_AVX512VNNI, F_AVX512BITALG): New.
23666 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
23668         PR target/83451
23669         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
23670         insn for floating-point loads and stores.
23672 2018-03-14  Carl Love  <cel@us.ibm.com>
23674         * config/rs6000/rs6000-c.c: Add macro definitions for
23675         ALTIVEC_BUILTIN_VEC_PERMXOR.
23676         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
23677         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
23678         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
23679         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
23680         UNSPEC_VPERMXOR.
23681         * config/doc/extend.texi: Add prototypes for vec_permxor.
23683 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
23685         PR c/84852
23686         * diagnostic-show-locus.c (class layout_point): Convert m_line
23687         from int to linenum_type.
23688         (line_span::comparator): Use linenum "compare" function when
23689         comparing line numbers.
23690         (test_line_span): New function.
23691         (layout_range::contains_point): Convert param "row" from int to
23692         linenum_type.
23693         (layout_range::intersects_line_p): Likewise.
23694         (layout::will_show_line_p): Likewise.
23695         (layout::print_source_line): Likewise.
23696         (layout::should_print_annotation_line_p): Likewise.
23697         (layout::print_annotation_line): Likewise.
23698         (layout::print_leading_fixits): Likewise.
23699         (layout::annotation_line_showed_range_p): Likewise.
23700         (struct line_corrections): Likewise for field m_row.
23701         (line_corrections::line_corrections): Likewise for param "row".
23702         (layout::print_trailing_fixits): Likewise.
23703         (layout::get_state_at_point): Likewise.
23704         (layout::get_x_bound_for_row): Likewise.
23705         (layout::print_line): Likewise.
23706         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
23707         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
23708         * input.c (selftest::test_linenum_comparisons): New function.
23709         (selftest::input_c_tests): Call it.
23710         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
23711         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
23712         * selftest.h (ASSERT_GT): New macro.
23713         (ASSERT_GT_AT): New macro.
23714         (ASSERT_LT): New macro.
23715         (ASSERT_LT_AT): New macro.
23717 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
23719         PR rtl-optimization/84780
23720         * combine.c (distribute_links): Don't make a link based on pc_rtx.
23722 2018-03-14  Martin Liska  <mliska@suse.cz>
23724         * tree.c (record_node_allocation_statistics): Use
23725         get_stats_node_kind.
23726         (get_stats_node_kind): New function extracted from
23727         record_node_allocation_statistics.
23728         (free_node): Use get_stats_node_kind.
23730 2018-03-14  Richard Biener  <rguenther@suse.de>
23732         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
23733         that the value-set of ANTIC_IN doesn't grow.
23735         Revert
23736         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
23737         member.
23738         (BB_VISITED_WITH_VISITED_SUCCS): New define.
23739         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
23741 2018-03-14  Julia Koval  <julia.koval@intel.com>
23743         * config.gcc (icelake-client, icelake-server): New.
23744         (icelake): Remove.
23745         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
23746         (initial_ix86_arch_features): Ditto.
23747         (PTA_SKYLAKE): Add SGX.
23748         (PTA_ICELAKE): Remove.
23749         (PTA_ICELAKE_CLIENT): New.
23750         (PTA_ICELAKE_SERVER): New.
23751         (ix86_option_override_internal): Split up icelake on icelake client and
23752         icelake server.
23753         (get_builtin_code_for_version): Ditto.
23754         (fold_builtin_cpu): Ditto.
23755         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
23756         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
23757         * config/i386/i386.h (processor_type): Ditto.
23758         * doc/invoke.texi: Ditto.
23760 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
23762         PR sanitizer/83392
23763         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
23764         INTEGER_CST offset, add it together with bitpos / 8 and
23765         sign extend based on POINTER_SIZE.
23767         PR target/84844
23768         Revert
23769         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
23771         PR target/78090
23772         * config/i386/constraints.md (Yc): New register constraint.
23773         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
23774         Use Yc constraint for alternative 2 of operand 0.  Remove
23775         preferred_for_speed attribute.
23777 2018-03-14  Richard Biener  <rguenther@suse.de>
23779         PR tree-optimization/84830
23780         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
23781         with the old one to avoid oscillations.
23783 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
23785         PR target/83712
23786         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
23787         pseudos.
23788         (assign_by_spills): Return a flag of reload assignment failure.
23789         Do not process the reload assignment failures.  Do not spill other
23790         reload pseudos if they has the same reg class.  Update n if
23791         necessary.
23792         (lra_assign): Add a return arg.  Set up from the result of
23793         assign_by_spills call.
23794         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
23795         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
23796         usage_insns if it is not NULL.
23797         (spill_hard_reg_in_range): New function.
23798         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
23799         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
23800         function prototypes.
23801         (lra_assign): Change prototype.
23802         * lra.c (lra): Add code to deal with fails by splitting hard reg
23803         live ranges.
23805 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
23807         * config/riscv/riscv.opt (mrelax): New option.
23808         * config/riscv/riscv.c (riscv_file_start): Emit ".option
23809         "norelax" when riscv_mrelax is disabled.
23810         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
23812 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
23814         PR target/84743
23815         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
23816         reassociation for int modes.
23818 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23820         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
23821         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
23822         for big-endian.
23823         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
23824         * config/aarch64/aarch64-sve.md
23825         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
23826         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
23827         (*extend<mode><Vwide>2): Rename to...
23828         (aarch64_sve_extend<mode><Vwide>2): ...this.
23829         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
23830         renaming the old pattern to...
23831         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
23832         unsigned packs.
23833         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
23834         define_expand, renaming the old pattern to...
23835         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
23836         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
23837         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
23838         account when deciding which SVE instruction the optab should use.
23839         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
23841 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23843         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
23844         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
23845         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
23846         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
23847         (tlsdesc_small_<mode>): Turn a define_expand and use
23848         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
23849         (tlsdesc_small_advsimd_<mode>): ...this.
23850         (tlsdesc_small_sve_<mode>): New pattern.
23852 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23854         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
23855         (UNSPEC_UMUL_HIGHPART): New constants.
23856         (MUL_HIGHPART): New int iteraor.
23857         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
23858         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
23859         define_expand.
23860         (*<su>mul<mode>3_highpart): New define_insn.
23862 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
23864         PR lto/84805
23865         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
23866         incomplete types.
23868 2018-03-13  Martin Liska  <mliska@suse.cz>
23870         PR ipa/84658.
23871         * (sem_item_optimizer::sem_item_optimizer): Initialize new
23872         vector.
23873         (sem_item_optimizer::~sem_item_optimizer): Release it.
23874         (sem_item_optimizer::merge_classes): Register variable aliases.
23875         (sem_item_optimizer::fixup_pt_set): New function.
23876         (sem_item_optimizer::fixup_points_to_sets): Likewise.
23877         * ipa-icf.h: Declare new variables and functions.
23879 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
23881         PR middle-end/84834
23882         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
23883         integer_pow2p@2 and test integer_pow2p in condition.
23884         (A < 0 ? C : 0): Similarly for @1.
23886         PR middle-end/84831
23887         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
23888         characters starting at p contain '\0' character, don't look beyond
23889         that.
23891         PR target/84827
23892         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
23893         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
23895         PR target/84828
23896         * reg-stack.c (change_stack): Change update_end var from int to
23897         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
23898         also call set_block_for_insn on the newly added insns and rescan.
23900         PR target/84786
23901         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
23902         on the last operand.
23904         PR c++/84704
23905         * tree.c (stabilize_reference_1): Return save_expr (e) for
23906         STATEMENT_LIST even if it doesn't have side-effects.
23908 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
23910         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
23912 2018-03-12  Renlin Li  <renlin.li@arm.com>
23914         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
23915         aarch64_output_scalar_simd_mov_immediate.
23917 2018-03-12  Martin Sebor  <msebor@redhat.com>
23919         PR tree-optimization/83456
23920         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
23921         for perfectly overlapping calls to memcpy.
23922         (gimple_fold_builtin_memory_chk): Same.
23923         (gimple_fold_builtin_strcpy): Handle no-warning.
23924         (gimple_fold_builtin_stxcpy_chk): Same.
23925         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
23927 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
23929         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
23930         parameter.  Use it for SFmode.
23931         (rs6000_function_arg_advance_1): Adjust.
23932         (rs6000_function_arg): Adjust.
23933         (rs6000_gimplify_va_arg): Pass false for that new parameter.
23935 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
23937         PR rtl-optimization/84169
23938         PR rtl-optimization/84780
23939         * combine.c (can_combine_p): Check for a 2-insn combination whether
23940         the destination register is used between the two insns, too.
23942 2018-03-12  Richard Biener  <rguenther@suse.de>
23944         PR tree-optimization/84803
23945         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
23946         for refs DR analysis didn't process.
23948 2018-03-12  Richard Biener  <rguenther@suse.de>
23950         PR tree-optimization/84777
23951         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
23952         force-vectorize loops ignore whether we are optimizing for size.
23954 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
23956         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
23957         (TARGET_MD_ASM_ADJUST): Define.
23959 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
23960             Kito Cheng  <kito.cheng@gmail.com>
23961             Chung-Ju Wu  <jasonwucj@gmail.com>
23963         * config/nds32/nds32.c (nds32_compute_stack_frame,
23964         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
23965         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
23966         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
23967         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
23968         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
23969         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
23970         * config/nds32/nds32.md (prologue, epilogue): Use macro
23971         NDS32_V3PUSH_AVAILABLE_P to do checking.
23973 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
23975         PR debug/58150
23976         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
23977         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
23978         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
23979         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
23980         addition of most attributes on !orig_type_die or the attribute not
23981         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
23983 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
23984             Chung-Ju Wu  <jasonwucj@gmail.com>
23986         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
23987         __NDS32_VH__ macro.
23988         * config/nds32/nds32.opt (mvh): New option.
23990 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
23991             Chung-Ju Wu  <jasonwucj@gmail.com>
23993         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
23994         function.
23995         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
23996         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
23997         definition.
23999 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
24000             Chung-Ju Wu  <jasonwucj@gmail.com>
24002         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
24003         function.
24004         * config/nds32/nds32-multiple.md (strlensi): New pattern.
24005         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
24007 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
24008             Kito Cheng  <kito.cheng@gmail.com>
24009             Chung-Ju Wu  <jasonwucj@gmail.com>
24011         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
24012         UNSPEC_FFMISM and UNSPEC_FLMISM.
24013         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
24014         for ffb, ffmism and flmism.
24015         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
24016         (unspec_ffmism): Ditto.
24017         (unspec_flmism): Ditto.
24018         (nds32_expand_builtin_impl): Check if string extension is available.
24019         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
24020         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
24022 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
24024         Reverting patch:
24025         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
24027         PR target/83712
24028         * lra-assigns.c (assign_by_spills): Return a flag of reload
24029         assignment failure.  Do not process the reload assignment
24030         failures.  Do not spill other reload pseudos if they has the same
24031         reg class.
24032         (lra_assign): Add a return arg.  Set up from the result of
24033         assign_by_spills call.
24034         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
24035         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
24036         usage_insns if it is not NULL.
24037         (spill_hard_reg_in_range): New function.
24038         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
24039         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
24040         function prototypes.
24041         (lra_assign): Change prototype.
24042         * lra.c (lra): Add code to deal with fails by splitting hard reg
24043         live ranges.
24045 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
24047         PR target/84807
24048         * config/i386/i386.opt: Replace Enforcment with Enforcement.
24050 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
24052         PR debug/84620
24053         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
24054         (dw_val_node): Add val_symbolic_view.
24055         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
24056         (symview_upper_bound): New.
24057         (new_line_info_table): Initialize symviews_since_reset.
24058         (dwarf2out_source_line): Count symviews_since_reset and set
24059         symview_upper_bound.
24060         (dw_val_equal_p): Handle symview.
24061         (add_AT_symview): New.
24062         (print_dw_val): Handle symview.
24063         (attr_checksum, attr_checksum_ordered): Likewise.
24064         (same_dw_val_p, size_of_die): Likewise.
24065         (value_format, output_die): Likewise.
24066         (add_high_low_attributes): Use add_AT_symview for entry_view.
24067         (dwarf2out_finish): Reset symview_upper_bound, clear
24068         zero_view_p.
24070 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
24072         PR target/83969
24073         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
24074         Add strict argument and use it.
24075         (rs6000_split_multireg_move): Update for new strict argument.
24076         (mem_operand_gpr): Disallow all non-offsettable addresses.
24077         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
24079 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
24081         PR target/84772
24082         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
24083         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
24084         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
24086         PR c++/84767
24087         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
24088         decl, use remap_type if we want to use the type.
24090 2018-03-09  Martin Sebor  <msebor@redhat.com>
24092         PR tree-optimization/84526
24093         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
24094         Remove dead code.
24095         (builtin_access::generic_overlap): Be prepared to handle non-array
24096         base objects.
24098 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
24100         PR rtl-optimization/84682
24101         * lra-constraints.c (process_address_1): Check is_address flag
24102         for address constraints.
24103         (process_alt_operands): Likewise.
24104         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
24105         preprocess_constraints.
24106         * recog.h (preprocess_constraints): Add oploc parameter.
24107         Adjust callers.
24108         * recog.c (preprocess_constraints): Test address_operand for
24109         CT_ADDRESS constraints.
24111 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
24113         PR target/83712
24114         * lra-assigns.c (assign_by_spills): Return a flag of reload
24115         assignment failure.  Do not process the reload assignment
24116         failures.  Do not spill other reload pseudos if they has the same
24117         reg class.
24118         (lra_assign): Add a return arg.  Set up from the result of
24119         assign_by_spills call.
24120         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
24121         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
24122         usage_insns if it is not NULL.
24123         (spill_hard_reg_in_range): New function.
24124         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
24125         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
24126         function prototypes.
24127         (lra_assign): Change prototype.
24128         * lra.c (lra): Add code to deal with fails by splitting hard reg
24129         live ranges.
24131 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24133         PR target/83193
24134         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
24135         Accept complain bool parameter.  Only emit errors if it is true.
24136         (arm_parse_cpu_option_name): Likewise.
24137         (arm_target_thumb_only): Adjust callers of the above.
24138         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
24139         prototype to take a default true bool parameter.
24140         (arm_parse_arch_option_name): Likewise.
24142 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
24143             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
24145         PR jit/64089
24146         PR jit/84288
24147         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
24148         * configure: Regenerate.
24149         * configure.ac ("linker --version-script option"): New.
24150         ("linker soname option"): New.
24152 2018-03-09  Richard Biener  <rguenther@suse.de>
24154         PR tree-optimization/84775
24155         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
24156         immediate uses of predicate stmts and mark them modified.
24158         Revert
24159         PR tree-optimization/84178
24160         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
24161         to caller.
24162         (version_loop_for_if_conversion): Delay update_ssa call.
24163         (tree_if_conversion): Delay update_ssa until after predicate
24164         insertion.
24166 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
24168         PR target/84763
24169         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
24170         when the function accesses prior frames.
24172 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
24174         PR debug/84456
24175         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
24176         gen_llsym, otherwise call maybe_gen_llsym.
24178         PR inline-asm/84742
24179         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
24180         has ',' character inside of it.
24182 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24184         PR target/84748
24185         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
24186         as clobbering CC_REGNUM.
24188 2018-03-08  Richard Biener  <rguenther@suse.de>
24190         PR middle-end/84552
24191         * tree-scalar-evolution.c: Include tree-into-ssa.h.
24192         (follow_copies_to_constant): Do not follow SSA names registered
24193         for update.
24195 2018-03-08  Richard Biener  <rguenther@suse.de>
24197         PR tree-optimization/84178
24198         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
24199         to caller.
24200         (version_loop_for_if_conversion): Delay update_ssa call.
24201         (tree_if_conversion): Delay update_ssa until after predicate
24202         insertion.
24204 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
24206         PR tree-optimization/84178
24207         * tree-if-conv.c (release_bb_predicate): Remove the
24208         the assertion that the stmts have NULL use_ops.
24209         Discard the statements, asserting that they haven't
24210         yet been added to a BB.
24212 2018-03-08  Richard Biener  <rguenther@suse.de>
24214         PR tree-optimization/84746
24215         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
24216         (phi_translate): Pass in destination ANTIC_OUT set.
24217         (phi_translate_1): Likewise.  For a simplified result lookup
24218         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
24219         (phi_translate_set): Adjust.
24220         (do_pre_regular_insertion): Likewise.
24221         (do_pre_partial_partial_insertion): Likewise.
24223 2018-03-08  Martin Liska  <mliska@suse.cz>
24225         PR gcov-profile/84735
24226         * doc/gcov.texi: Document usage of profile files.
24227         * gcov-io.h: Document changes in the format.
24229 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
24231         PR debug/84404
24232         PR debug/84408
24233         * dwarf2out.c (struct dw_line_info_table): Update comments for
24234         view == -1.
24235         (FORCE_RESET_NEXT_VIEW): New.
24236         (FORCE_RESETTING_VIEW_P): New.
24237         (RESETTING_VIEW_P): Check for -1 too.
24238         (ZERO_VIEW_P): Likewise.
24239         (new_line_info_table): Force-reset next view.
24240         (dwarf2out_begin_function): Likewise.
24241         (dwarf2out_source_line): Simplify zero_view_p initialization.
24242         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
24243         view directly.  Omit view when omitting .loc at line 0.
24245 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
24247         PR tree-optimization/84740
24248         * tree-switch-conversion.c (process_switch): Call build_constructors
24249         only if info.phi_count is non-zero.
24251         PR tree-optimization/84739
24252         * tree-tailcall.c (find_tail_calls): Check call arguments against
24253         DECL_ARGUMENTS (current_function_decl) rather than
24254         DECL_ARGUMENTS (func) when checking for tail recursion.
24256 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
24258         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
24259         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
24260         Volker Reichelt's entry and add entries for people that perform
24261         GCC fuzzy testing and report numerous bugs.
24263 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
24265         PR target/82411
24266         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
24267         readonly data in sdata, if that is disabled.
24268         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
24269         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
24270         -mreadonly-in-sdata option.
24272 2018-03-07  Martin Sebor  <msebor@redhat.com>
24274         PR tree-optimization/84468
24275         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
24276         basic block when looking for nul assignment.
24278 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
24280         PR target/84277
24281         * except.h (output_function_exception_table): Adjust prototype.
24282         * except.c (output_function_exception_table): Remove FNNAME parameter
24283         and add SECTION parameter.  Ouput one part of the table at a time.
24284         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
24285         the first part of the exception table and emit unwind directives.
24286         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
24287         (i386_pe_seh_cold_init): Likewise.
24288         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
24289         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
24290         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
24291         (ix86_output_call_insn): Emit a nop in one more case for SEH.
24292         * config/i386/winnt.c: Include except.h.
24293         (struct seh_frame_state): Add reg_offset, after_prologue and
24294         in_cold_section fields.
24295         (i386_pe_seh_end_prologue): Set seh->after_prologue.
24296         (i386_pe_seh_cold_init): New function.
24297         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
24298         to seh->in_cold_section.
24299         (seh_emit_push): Record the offset of the push.
24300         (seh_emit_save): Record the offet of the save.
24301         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
24302         Test seh->after_prologue to disregard the epilogue.
24303         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
24304         (i386_pe_end_cold_function): New function.
24306 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
24308         PR fortran/84565
24309         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
24310         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
24312         PR c++/84704
24313         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
24314         on tmp_var.
24315         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
24316         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
24318         PR middle-end/84723
24319         * multiple_target.c: Include tree-inline.h and intl.h.
24320         (expand_target_clones): Diagnose and fail if node->definition and
24321         !tree_versionable_function_p (node->decl).
24323 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
24325         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
24326         sprint_ul.
24327         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
24328         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
24329         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
24331 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
24333         PR target/84710
24334         * combine.c (try_combine): Use reg_or_subregno instead of handling
24335         just paradoxical SUBREGs and REGs.
24337 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
24339         * config/arc/arc.c (arc_finalize_pic): Remove function.
24340         (arc_must_save_register): We use single base PIC register, remove
24341         checks to save/restore the PIC register.
24342         (arc_expand_prologue): Likewise.
24343         * config/arc/arc-protos.h (arc_set_default_type_attributes):
24344         Remove.
24345         (arc_verify_short): Likewise.
24346         (arc_attr_type): Likewise.
24347         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
24348         (walk_stores): Likewise.
24349         (arc_address_cost): Make it static.
24350         (arc_verify_short): Likewise.
24351         (branch_dest): Likewise.
24352         (arc_attr_type): Likewise.
24353         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
24354         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
24355         (arc_final_prescan_insn): Remove inserting the nops due to
24356         hardware hazards.  It is done in reorg step.
24357         (insn_length_variant_t): Remove.
24358         (insn_length_parameters_t): Likewise.
24359         (arc_insn_length_parameters): Likewise.
24360         (arc_get_insn_variants): Likewise.
24361         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
24363 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
24365         PR inline-asm/84683
24366         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
24367         assertion failure.
24369         PR tree-optimization/84687
24370         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
24371         on new_node->decl.
24372         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
24374 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24376         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
24377         Rename to ppc_speculation_barrier.
24378         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
24379         __builtin_ppc_speculation_barrier.
24381 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
24383         PR target/84700
24384         * combine.c (combine_simplify_rtx): Don't try to simplify if
24385         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
24386         are equal to x.
24388 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
24390         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
24391         to 32 bytes when compiling for POWER9.
24393 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
24395         PR target/84564
24396         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
24397         regparm >= 3 with no arg reg available also for calls with
24398         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
24400         PR target/84524
24401         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
24402         orig,vex.
24403         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
24405 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
24407         PR target/84264
24408         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
24410 2018-03-05  Richard Biener  <rguenther@suse.de>
24412         PR tree-optimization/84486
24413         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
24414         When inserting a __builtin_assume_aligned call set the LHS
24415         SSA name alignment info accordingly.
24417 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
24419         PR tree-optimization/84114
24420         * config/aarch64/aarch64.c (aarch64_reassociation_width)
24421         Avoid reassociation of FLOAT_MODE addition.
24423 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
24425         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
24426         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
24427         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
24428         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
24429         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
24430         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
24431         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
24432         and -mwbnoinvd.
24433         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
24434         __builtin_ia32_wbinvd): New builtins.
24435         (SPECIAL_ARGS2): New.
24436         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
24437         (SPECIAL_ARGS2): New.
24438         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
24439         (ix86_valid_target_attribute_inner_p): Ditto.
24440         (ix86_init_mmx_sse_builtins): Add special_args2.
24441         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
24442         TARGET_WBNOINVD_P): New.
24443         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
24444         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
24445         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
24446         * config/i386/immintrin.h (_wbinvd): New intrinsic.
24447         * config/i386/pconfigintrin.h: New file.
24448         * config/i386/wbnoinvdintrin.h: Ditto.
24449         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
24450         wbnoinvdintrin.h.
24451         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
24453 2018-03-05  Richard Biener  <rguenther@suse.de>
24455         PR tree-optimization/84670
24456         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
24457         member.
24458         (BB_VISITED_WITH_VISITED_SUCCS): New define.
24459         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
24460         (compute_antic_aux): Only assert the number of values in ANTIC_IN
24461         doesn't grow if all successors (recursively) were visited at least
24462         once.
24464 2018-03-05  Richard Biener  <rguenther@suse.de>
24466         PR tree-optimization/84650
24467         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
24468         if executed in the loop pipeline.
24470 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
24472         * doc/configfiles.texi (Configuration Files): Move info about
24473         conditionalizing $target-protos.h to...
24474         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
24475         differs from $target-protos.h.
24477 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
24478             Chung-Ju Wu  <jasonwucj@gmail.com>
24480         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
24481         * config/nds32/nds32-multiple.md (setmemsi): Define.
24482         * config/nds32/nds32-memory-manipulation.c
24483         (nds32_gen_dup_4_byte_to_word_value): New.
24484         (emit_setmem_word_loop): New.
24485         (emit_setmem_byte_loop): New.
24486         (nds32_expand_setmem_loop): New.
24487         (nds32_expand_setmem_loop_v3m): New.
24488         (nds32_expand_setmem_unroll): New.
24489         (nds32_expand_setmem): New.
24491 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24492             Chung-Ju Wu  <jasonwucj@gmail.com>
24494         * config/nds32/nds32-memory-manipulation.c
24495         (nds32_emit_load_store): New.
24496         (nds32_emit_post_inc_load_store): New.
24497         (nds32_emit_mem_move): New.
24498         (nds32_emit_mem_move_block): New.
24499         (nds32_expand_movmemsi_loop_unknown_size): New.
24500         (nds32_expand_movmemsi_loop_known_size): New.
24501         (nds32_expand_movmemsi_loop): New.
24502         (nds32_expand_movmemsi_unroll): New.
24503         (nds32_expand_movmemqi): Rename ...
24504         (nds32_expand_movmemsi): ... to this.
24505         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
24506         (movmemsi): ... to this.
24507         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
24508         (nds32_expand_movmemsi): ... to this.
24510 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24511             Monk Chiang  <sh.chiang04@gmail.com>
24512             Chung-Ju Wu  <jasonwucj@gmail.com>
24514         * config/nds32/nds32-protos.h
24515         (nds32_expand_load_multiple): New arguments.
24516         (nds32_expand_store_multiple): Ditto.
24517         (nds32_valid_multiple_load_store): Rename ...
24518         (nds32_valid_multiple_load_store_p): ... to this.
24519         * config/nds32/nds32-memory-manipulation.c
24520         (nds32_expand_load_multiple): Refine implementation.
24521         (nds32_expand_store_multiple): Ditto.
24522         * config/nds32/nds32-multiple.md
24523         (load_multiple): Update nds32_expand_load_multiple interface.
24524         (store_multiple): Update nds32_expand_store_multiple interface.
24525         * config/nds32/nds32-predicates.c
24526         (nds32_valid_multiple_load_store): Rename ...
24527         (nds32_valid_multiple_load_store_p): ... to this and refine
24528         implementation.
24529         * config/nds32/predicates.md
24530         (nds32_load_multiple_and_update_address_operation): New predicate.
24531         (nds32_store_multiple_and_update_address_operation): New predicate.
24533 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24534             Chung-Ju Wu  <jasonwucj@gmail.com>
24536         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
24537         (combo): New attribute.
24538         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
24540 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
24542         * config/nds32/nds32.opt: Change -mcmodel= default value.
24544 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
24545             Monk Chiang  <sh.chiang04@gmail.com>
24546             Chung-Ju Wu  <jasonwucj@gmail.com>
24548         * config/nds32/constants.md (unspec_element): New enum.
24549         * config/nds32/constraints.md (Umw): New constraint.
24550         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
24551         * config/nds32/nds32-intrinsic.md: Likewise.
24552         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
24553         (nds32_valid_smw_lwm_base_p): New.
24554         (nds32_output_smw_single_word): New.
24555         (nds32_output_lmw_single_word): New.
24556         (nds32_expand_unaligned_load): New.
24557         (nds32_expand_unaligned_store): New.
24558         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
24559         (nds32_output_smw_single_word): Declare.
24560         (nds32_output_lmw_single_word): Declare.
24561         (nds32_expand_unaligned_load): Declare.
24562         (nds32_expand_unaligned_store): Declare.
24563         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
24564         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
24565         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
24566         NDS32_BUILTIN_UASTORE_DW.
24567         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
24568         predicate.
24570 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
24571             Kito Cheng  <kito.cheng@gmail.com>
24572             Chung-Ju Wu  <jasonwucj@gmail.com>
24574         * config/nds32/nds32-intrinsic.c
24575         (nds32_expand_builtin_null_ftype_reg): Delete.
24576         (nds32_expand_builtin_reg_ftype_imm): Ditto.
24577         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
24578         (nds32_read_argument): New.
24579         (nds32_legitimize_target): Ditto.
24580         (nds32_legitimize_argument): Ditto.
24581         (nds32_check_constant_argument): Ditto.
24582         (nds32_expand_unop_builtin): Ditto.
24583         (nds32_expand_unopimm_builtin): Ditto.
24584         (nds32_expand_binop_builtin): Ditto.
24585         (nds32_builtin_decl_impl): Ditto.
24586         (builtin_description): Ditto.
24587         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
24588         (nds32_init_builtins_impl): Ditto.
24589         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
24590         (nds32_builtin_decl): New.
24591         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
24592         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
24594 2018-03-02  Jeff Law  <law@redhat.com>
24596         * reorg.c (stop_search_p): Handle DEBUG_INSN.
24597         (redundant_insn, fill_simple_delay_slots): Likewise.
24598         (fill_slots_from_thread): Likewise.
24599         * resource.c (mark_referenced_resources): Likewise.
24600         (mark_set_resources, find_dead_or_set_registers): Likewise.
24602 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24604         * substring-locations.h (format_warning_va): Formatting fix for
24605         ATTRIBUTE_GCC_DIAG.
24606         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
24607         argument.
24608         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
24609         * substring-locations.c: Include intl.h.
24610         (format_warning_va): Turned into small wrapper around
24611         format_warning_n_va, renamed to ...
24612         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
24613         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
24614         use ngettext.
24615         (format_warning_at_substring_n): New function.
24616         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
24617         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
24618         format_warning_at_substring with just a shorter name instead of
24619         const function pointer.
24620         (fmtwarn_n): New function.
24621         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
24622         appropriate, get rid of all the fmtstr temporaries, move conditionals
24623         with G_() wrapped string literals directly into fmtwarn arguments,
24624         cast dir.len to (int), formatting fixes.
24626 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
24628         * doc/invoke.texi: Remove "Cilk Plus" references.
24630 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24631             Richard Biener  <rguenther@suse.de>
24633         PR ipa/84628
24634         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
24635         for error or warning attributes if CALL_FROM_THUNK_P is set.
24636         Formatting fixes.
24638 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24640         PR target/56540
24641         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
24642         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
24644         PR target/56540
24645         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
24646         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
24648         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
24649         instead of -1U in last predictors element's probability member.
24651 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
24653         PR ipa/83983
24654         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
24655         arguments if they are comparable.
24657 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
24659         PR tree-optimization/84634
24660         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
24661         masks and masked_loop_p with a single loop_masks, making sure it's
24662         null for bb vectorization.
24664 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
24666         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
24667         (vect_analyze_data_ref_access): Use loop->safe_len rather than
24668         loop->force_vectorize to check whether there is no alias.
24670 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24672         PR target/84614
24673         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
24674         prototypes.
24675         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
24676         comments.
24677         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
24678         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
24679         instead of a loop around prev_real_insn.
24680         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
24681         prev_real_insn.
24683         PR inline-asm/84625
24684         * config/i386/i386.c (ix86_print_operand): Use conditional
24685         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
24686         zero vector.
24688 2018-03-02  Richard Biener  <rguenther@suse.de>
24690         PR tree-optimization/84427
24691         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
24692         (bitmap_set_subtract_values): Rewrite to handle multiple
24693         exprs per value.
24694         (clean): Likewise.
24695         (prune_clobbered_mems): Likewise.
24696         (phi_translate): Take edge instead of pred/phiblock.
24697         (phi_translate_1): Likewise.
24698         (phi_translate_set): Likewise.  Insert all translated
24699         exprs for a value into the set, keeping possibly multiple
24700         expressions per value.
24701         (compute_antic_aux): Adjust for phi_translate changes.
24702         When intersecting union the expressions and prune those
24703         not in the final value set, keeping possibly multiple
24704         expressions per value.  Do not use value-insertion
24705         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
24706         all expressions.  Add verification that the value-sets
24707         only shrink during iteration.
24708         (compute_partial_antic_aux): Adjust for the phi_translate changes.
24709         (do_pre_regular_insertion): Likewise.
24710         (do_pre_partial_partial_insertion): Likewise.
24712 2018-03-02  Richard Biener  <rguenther@suse.de>
24714         PR target/82005
24715         * config/darwin.c (saved_debug_info_level): New static global.
24716         (darwin_asm_lto_start): Disable debug info generation for LTO out.
24717         (darwin_asm_lto_end): Restore debug info generation settings.
24719 2018-03-01  Martin Liska  <mliska@suse.cz>
24721         PR sanitizer/82484
24722         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
24723         volatile arguments.
24725 2018-03-01  Richard Biener  <rguenther@suse.de>
24727         PR debug/84645
24728         * dwarf2out.c (gen_variable_die): Properly handle late VLA
24729         type annotation with LTO when debug was disabled at compile-time.
24731 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
24733         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
24734         XINT with INTVAL.
24735         (mips_final_postscan_insn): Likewise.
24737 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
24739         PR rtl-optimization/84528
24740         * alias.c (init_alias_target): Add commentary.
24741         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
24742         a unique base value if the frame pointer is not eliminated
24743         to the stack pointer.
24745 2018-03-01  Tom de Vries  <tom@codesourcery.com>
24747         PR rtl-optimization/83327
24748         * lra-int.h (hard_regs_spilled_into): Declare.
24749         * lra.c (hard_regs_spilled_into): Define.
24750         (init_reg_info): Init hard_regs_spilled_into.
24751         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
24752         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
24753         (process_bb_lives): Handle hard_regs_spilled_into.
24754         (lra_create_live_ranges_1): Before doing liveness propagation, clear
24755         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
24757 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
24759         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
24760         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
24761         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
24762         * config/rs6000/aix72.h: New file.
24764 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
24766         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
24767         instead of warning_at with conditional singular and plural messages
24768         where possible.
24770         PR target/52991
24771         * stor-layout.c (update_alignment_for_field): For
24772         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
24773         && !DECL_PACKED (field), do the alignment update, just use
24774         only desired_align instead of MAX (type_align, desired_align)
24775         as the alignment.
24776         (place_field): Don't do known_align < desired_align handling
24777         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
24778         is non-NULL, instead do it after rli->prev_field handling and
24779         only if not within a bitfield word.  For DECL_PACKED (field)
24780         use type_align of BITS_PER_UNIT.
24782 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
24784         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
24785         superfluous parentheses and trailing spaces.
24787 2018-02-28  Richard Biener  <rguenther@suse.de>
24789         PR tree-optimization/84584
24790         * graphite-scop-detection.c (scop_detection::add_scop): Discard
24791         SCoPs with fake exit edge.
24793 2018-02-28  Martin Liska  <mliska@suse.cz>
24795         PR testsuite/84597
24796         * timevar.c (timer::print): Fix format to properly print 100%
24797         values.
24799 2018-02-28  Richard Biener  <rguenther@suse.de>
24801         PR middle-end/84607
24802         * genmatch.c (capture_info::walk_match): Do not mark
24803         captured expressions without operands as expr_p given
24804         they act more like predicates and should be subject to
24805         "lost tail" side-effect preserving.
24807 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
24809         PR rtl-optimization/81611
24810         * auto-inc-dec.c (attempt_change): Move dead note from
24811         mem_insn if it's the next use of regno
24812         (find_address): Take address use of reg holding
24813         non-incremented value.  Add parm to limit search to the named
24814         reg only.
24815         (merge_in_block): Attempt to use a mem insn that is the next
24816         use of the original regno.
24818 2018-02-27  Martin Sebor  <msebor@redhat.com>
24820         PR c++/83871
24821         * doc/invoke.texi (-Wmissing-attributes): New option.
24822         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
24824 2018-02-27  Martin Sebor  <msebor@redhat.com>
24826         PR translation/84207
24827         * diagnostic-core.h (warning_n, error_n, inform_n): Change
24828         n argument to unsigned HOST_WIDE_INT.
24829         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
24830         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
24831         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
24832         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
24834 2018-02-27  Richard Biener  <rguenther@suse.de>
24836         PR tree-optimization/84512
24837         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
24838         Do not use the estimate returned from record_stmt_cost for
24839         the scalar iteration cost but sum properly using add_stmt_cost.
24841 2018-02-27  Richard Biener  <rguenther@suse.de>
24843         PR tree-optimization/84466
24844         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
24845         Adjust last change to less strictly validate use operands.
24847 2018-02-27  Martin Liska  <mliska@suse.cz>
24849         PR gcov-profile/84548
24850         * gcov.c (process_file): Allow partial overlap and consider it
24851         also as group functions.
24852         (output_lines): Properly calculate range of lines for a group.
24854 2018-02-27  Martin Liska  <mliska@suse.cz>
24856         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
24857         'ggc' suffixes.  Change first column width.
24858         (timer::print): Fix formatting of the column.
24860 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
24862         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
24863         preserve inline entry blocks for the sake of debug inline
24864         entry point markers alone.
24865         (remove_unused_locals): Suggest in comments a better place to
24866         force the preservation of inline entry blocks that are
24867         otherwise unused, but do not preserve them.
24869 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24871         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
24873 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24875         PR target/84039
24876         * config/i386/constraints.md (Bs): Replace
24877         ix86_indirect_branch_register with
24878         TARGET_INDIRECT_BRANCH_REGISTER.
24879         (Bw): Likewise.
24880         * config/i386/i386.md (indirect_jump): Likewise.
24881         (tablejump): Likewise.
24882         (*sibcall_memory): Likewise.
24883         (*sibcall_value_memory): Likewise.
24884         Peepholes of indirect call and jump via memory: Likewise.
24885         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
24886         (*sibcall_value_GOT_32): Likewise.
24887         * config/i386/predicates.md (indirect_branch_operand): Likewise.
24888         (GOT_memory_operand): Likewise.
24889         (call_insn_operand): Likewise.
24890         (sibcall_insn_operand): Likewise.
24891         (GOT32_symbol_operand): Likewise.
24892         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
24894 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
24896         PR rtl-optimization/83496
24897         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
24898         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
24899         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
24900         redundant insn, if any.
24901         (relax_delay_slots): Likewise.
24902         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
24904 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
24906         PR tree-optimization/83965
24907         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
24908         that grouped statements are part of a reduction chain.  Return
24909         true if the statement is not marked as a reduction itself but
24910         is part of a group.
24911         (vect_recog_dot_prod_pattern): Don't check whether the statement
24912         is part of a group here.
24913         (vect_recog_sad_pattern): Likewise.
24914         (vect_recog_widen_sum_pattern): Likewise.
24916 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
24918         PR debug/84545
24919         * final.c (rest_of_clean_state): Also look for calls inside sequences.
24921 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24923         PR target/84530
24924         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
24925         the bool argument.
24926         (ix86_output_indirect_function_return): New prototype.
24927         (ix86_split_simple_return_pop_internal): Likewise.
24928         * config/i386/i386.c (indirect_return_via_cx): New.
24929         (indirect_return_via_cx_bnd): Likewise.
24930         (indirect_thunk_name): Handle return va CX_REG.
24931         (output_indirect_thunk_function): Create alias for
24932         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
24933         (ix86_output_indirect_jmp): Remove the bool argument.
24934         (ix86_output_indirect_function_return): New function.
24935         (ix86_split_simple_return_pop_internal): Likewise.
24936         * config/i386/i386.md (*indirect_jump): Don't pass false
24937         to ix86_output_indirect_jmp.
24938         (*tablejump_1): Likewise.
24939         (simple_return_pop_internal): Change it to define_insn_and_split.
24940         Call ix86_split_simple_return_pop_internal to split it for
24941         -mfunction-return=.
24942         (simple_return_indirect_internal): Call
24943         ix86_output_indirect_function_return instead of
24944         ix86_output_indirect_jmp.
24946 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
24948         PR bootstrap/84405
24949         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
24950         memset and value initialization afterwards.
24952 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
24954         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
24956 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
24958         PR target/84521
24959         * common/config/aarch64/aarch64-common.c
24960         (aarch_option_optimization_table[]): Switch
24961         off fomit-frame-pointer
24963 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
24964             Chung-Ju Wu  <jasonwucj@gmail.com>
24966         * config/nds32/nds32-multiple.md (load_multiple): Disallow
24967         volatile memory.
24968         (store_multiple): Ditto.
24970 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
24972         * config.gcc: Add --with-cpu support for nds32 target.
24973         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
24974         * config/nds32/nds32.opt: Add -mcpu= option.
24976 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
24978         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
24979         isel=yes): Warn for these deprecated options.
24981 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
24983         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
24984         ISA_2_5_MASKS_EMBEDDED.
24986 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
24988         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
24989         p->max as pointers rather than using iterative_hash_expr.
24991 2018-02-23  Carl Love  <cel@us.ibm.com>
24993         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
24994         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
24995         BU_P8V_OVERLOAD_2.
24996         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
24997         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
24998         P8V_BUILTIN_VEC_VUNSIGNED2.
25000 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
25002         PR target/81572
25003         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
25004         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
25005         LRA_UNKNOWN_ALT.
25006         * lra-constraints.c (curr_insn_transform): Set up
25007         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
25008         LRA_UNKNOWN_ALT.
25009         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
25010         * lra-eliminations.c (spill_pseudos): Ditto.
25011         (process_insn_for_elimination): Ditto.
25012         * lra-lives.c (reg_early_clobber_p): Use the new macros.
25013         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
25014         LRA_NON_CLOBBERED_ALT.
25016 2018-02-22  Martin Sebor  <msebor@redhat.com>
25018         PR tree-optimization/84480
25019         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
25020         to maybe_diag_stxncpy_trunc.  Call it.
25021         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
25022         from gimple_fold_builtin_strcpy.  Print inlining stack.
25023         (handle_builtin_stxncpy): Print inlining stack.
25024         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
25026 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
25028         PR target/84176
25029         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
25030         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
25031         and -fcheck-pointer-bounds are used together.
25032         (indirect_thunk_prefix): New enum.
25033         (indirect_thunk_need_prefix): New function.
25034         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
25035         "_nt" instead of "_bnd" for NOTRACK prefix.
25036         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
25037         (output_indirect_thunk_function): Likewise.
25038         (): Likewise.
25039         (ix86_code_end): Update output_indirect_thunk_function calls.
25040         (ix86_output_indirect_branch_via_reg): Replace
25041         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
25042         (ix86_output_indirect_branch_via_push): Likewise.
25043         (ix86_output_function_return): Likewise.
25044         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
25045         incompatible with -fcf-protection=branch and
25046         -fcheck-pointer-bounds.
25048 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
25050         PR target/83335
25051         * config/aarch64/aarch64.c (aarch64_print_address_internal):
25052         Change gcc_assert call to output_operand_lossage.
25054 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
25056         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
25058 2018-02-22  DJ Delorie  <dj@redhat.com>
25059             Sebastian Perta  <sebastian.perta@renesas.com>
25060             Oleg Endo  <olegendo@gcc.gnu.org>
25062         * config/rx/rx.c (rx_rtx_costs): New function.
25063         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
25065 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25067         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
25069 2018-02-22  Martin Liska  <mliska@suse.cz>
25071         PR driver/83193
25072         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
25073         Add "native" as a possible value.
25075 2018-02-22  Martin Liska  <mliska@suse.cz>
25077         PR driver/83193
25078         * config/i386/i386.c (ix86_option_override_internal):
25079         Add "native" as a possible value for -march and -mtune.
25081 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
25083         PR target/84502
25084         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
25085         to all type variants.
25087         PR tree-optimization/84503
25088         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
25089         width as info->bitpos + info->bitsize - start.
25090         (merged_store_group::merge_overlapping): Simplify width computation.
25091         (check_no_overlap): New function.
25092         (imm_store_chain_info::try_coalesce_bswap): Compute expected
25093         start + width and last_order of the group, fail if check_no_overlap
25094         fails.
25095         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
25096         to group if check_no_overlap fails.
25098 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
25100         * config/rs6000/altivec.md: Delete contraint arguments to
25101         define_expand, define_split, and define_peephole2, and in
25102         define_insn_and_split if always unused.
25103         * config/rs6000/darwin.md: Ditto.
25104         * config/rs6000/dfp.md: Ditto.
25105         * config/rs6000/rs6000.md: Ditto.
25106         * config/rs6000/sync.md: Ditto.
25107         * config/rs6000/vector.md: Ditto.
25108         * config/rs6000/vsx.md: Ditto.
25110 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
25112         * config/rs6000/altivec.md: Write output control strings as braced
25113         blocks instead of double-quoted strings.
25114         * config/rs6000/darwin.md: Ditto.
25115         * config/rs6000/rs6000.md: Ditto.
25116         * config/rs6000/vector.md: Ditto.
25117         * config/rs6000/vsx.md: Ditto.
25119 2018-02-21  Jason Merrill  <jason@redhat.com>
25121         PR c++/84314 - ICE with templates and fastcall attribute.
25122         * attribs.c (build_type_attribute_qual_variant): Remove assert.
25124 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
25126         * ipa-cp.c (determine_versionability): Fix comment typos.
25128 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
25130         PR c/84229
25131         * ipa-cp.c (determine_versionability): Do not version functions caling
25132         va_arg_pack.
25134 2018-02-21  Martin Liska  <mliska@suse.cz>
25136         PR driver/83193
25137         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
25138         Add "native" as a possible value.
25139         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
25140         the macro when native cpu detection is available.
25142 2018-02-21  Martin Liska  <mliska@suse.cz>
25144         PR driver/83193
25145         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
25146         Add "native" as a possible value.
25147         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
25148         when native cpu detection is available.
25150 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
25151             Martin Sebor  <msebor@redhat.com>
25153         PR tree-optimization/84478
25154         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
25155         false.
25156         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
25157         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
25158         support which is conservatively correct, for 2 only stay conservative
25159         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
25160         argument to the 2 argument get_range_strlen, adjust 6 arg
25161         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
25162         false.
25163         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
25164         (gimple_fold_builtin_strlen): Pass true as last argument to
25165         get_range_strlen.
25167 2018-02-20  Martin Sebor  <msebor@redhat.com>
25169         PR middle-end/84095
25170         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
25171         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
25172         (builtin_memref::builtin_memref): Factor out parts into
25173         set_base_and_offset and call it.
25175 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
25177         PR middle-end/84406
25178         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
25179         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
25180         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
25181         search at the associated MODE_INT.
25183 2018-02-20  Jeff Law  <law@redhat.com>
25185         PR middle-end/82123
25186         PR tree-optimization/81592
25187         PR middle-end/79257
25188         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
25189         for range data rather than using global data.
25190         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
25191         range data rather than using global data.
25192         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
25193         pass it to children as needed.
25194         (struct directive::fmtresult): Similarly.
25195         (struct directive::set_width): Similarly.
25196         (struct directive::set_precision): Similarly.
25197         (format_integer, format_directive, parse_directive): Similarly.
25198         (format_none): Accept unnamed vr_values parameter.
25199         (format_percent, format_floating, format_character): Similarly.
25200         (format_string, format_plain): Similarly.
25201         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
25202         the EVRP range analyzer for range data rather than using global data.
25203         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
25204         gimple-ssa-evrp-analyze.h
25205         (class sprintf_dom_walker): Add after_dom_children member function.
25206         Add evrp_range_analyzer member.
25207         (sprintf_dom_walker::before_dom_children): Call into the EVRP
25208         range analyzer as needed.
25209         (sprintf_dom_walker::after_dom_children): New member function.
25210         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
25211         if not optimizing.
25212         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
25213         (evrp_range_analyzer::pop_to_marker): Likewise.
25215 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
25217         PR tree-optimization/84419
25218         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
25219         with the required type if its current type is compatible but
25220         different.
25222 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
25224         PR middle-end/82004
25225         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
25226         after vectorization.
25228 2018-02-20  Martin Liska  <mliska@suse.cz>
25230         PR driver/83193
25231         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
25232         possible values if we don't have a hint.
25234 2018-02-20  Martin Liska  <mliska@suse.cz>
25236         PR c/84310
25237         PR target/79747
25238         * final.c (shorten_branches): Build align_tab array with one
25239         more element.
25240         * opts.c (finish_options): Add alignment option limit check.
25241         (MAX_CODE_ALIGN): Likewise.
25242         (MAX_CODE_ALIGN_VALUE): Likewise.
25243         * doc/invoke.texi: Document maximum allowed option value for
25244         all -falign-* options.
25246 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
25248         PR target/84146
25249         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
25250         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
25251         * var-tracking.c (emit_note_insn_var_location): Remove all references
25252         to NOTE_INSN_CALL_ARG_LOCATION.
25253         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
25254         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
25255         Use copy_rtx_if_shared.
25256         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
25257         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
25258         (dwarf2out_var_location): Remove handling of
25259         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
25260         on call_insn.
25261         * final.c (final_scan_insn): Remove all references to
25262         NOTE_INSN_CALL_ARG_LOCATION.
25263         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
25264         before dumping final insns.
25265         * except.c (emit_note_eh_region_end): Remove all references to
25266         NOTE_INSN_CALL_ARG_LOCATION.
25267         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
25268         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
25269         * config/arc/arc.c (hwloop_optimize): Likewise.
25270         * config/arm/arm.c (create_fix_barrier): Likewise.
25271         * config/s390/s390.c (s390_chunkify_start): Likewise.
25272         * config/sh/sh.c (find_barrier): Likewise.
25273         * config/i386/i386.c (rest_of_insert_endbranch,
25274         ix86_seh_fixup_eh_fallthru): Likewise.
25275         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
25276         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
25277         * config/frv/frv.c (frv_function_prologue): Likewise.
25278         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
25279         reg note.
25280         (note_outside_basic_block_p): Remove all references to
25281         NOTE_INSN_CALL_ARG_LOCATION.
25282         * gengtype.c (adjust_field_rtx_def): Likewise.
25283         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
25284         Likewise.
25285         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
25286         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
25288         PR c++/84444
25289         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
25290         is ADDR_EXPR.
25292         PR tree-optimization/84452
25293         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
25294         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
25295         is NULL.
25297 2018-02-19  Martin Liska  <mliska@suse.cz>
25299         PR sanitizer/82183
25300         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
25302 2018-02-19  Martin Liska  <mliska@suse.cz>
25303             Richard Sandiford  <richard.sandiford@linaro.org>
25305         PR tree-optimization/82491
25306         * gimple-fold.c (get_base_constructor): Make earlier bail out
25307         to prevent ubsan.
25309 2018-02-19  Carl Love  <cel@us.ibm.com>
25311         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
25312         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
25313         BU_P8V_OVERLOAD_1.
25314         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
25315         P8V_BUILTIN_VEC_NEG.
25317 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
25319         * config/rl78/rl78.md (movdf): New define expand.
25321 2018-02-19  Martin Liska  <mliska@suse.cz>
25323         PR other/80589
25324         * doc/invoke.texi: Fix typo.
25325         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
25327 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
25329         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
25330         handle rs6000_single_float and rs6000_double_float specially for
25331         e500 family CPUs.
25333 2018-02-16  Jeff Law  <law@redhat.com>
25335         * config/rx/rx.c (add_pop_cfi_notes): New function.;
25336         (pop_regs): Use it.
25338 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
25340         PR ipa/84425
25341         * ipa-inline.c (inline_small_functions): Fix a typo.
25343 2018-02-16  Nathan Sidwell  <nathan@acm.org>
25345         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
25347 2018-02-16  Carl Love  <cel@us.ibm.com>
25349         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
25350         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
25351         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
25352         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
25353         expansion to P8V_BUILTIN_VEC_FLOAT2.
25355 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
25357         PR rtl-optimization/70023
25358         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
25359         src_regno into account.
25361 2018-02-16  Carl Love  <cel@us.ibm.com>
25363         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
25364         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
25365         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
25366         * config/rs6000/rs6000.c: Remove case statements for
25367         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
25368         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
25369         and P9V_BUILTIN_VEC_VINSERT4B.
25370         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
25371         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
25372         * config/rs6000/vsx.md:
25373         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
25374         vec_insert4b.
25376 2018-02-16  Carl Love  <cel@us.ibm.com>
25378         * config/rs6000/altivec.h: Add builtin names vec_extract4b
25379         vec_insert4b.
25380         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
25381         definitions.
25382         * config/rs6000/rs6000-c.c: Add the definitions for
25383         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
25384         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
25385         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
25386         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
25387         definition for insert4b and define insn *insert3b_internal.
25388         * doc/extend.texi: Add documentation for vec_extract4b.
25390 2018-02-16  Nathan Sidwell  <nathan@acm.org>
25392         * doc/extend.texi (Backwards Compatibility): Mention friend
25393         injection.  Note for-scope is deprecated.
25394         * doc/invoke.texi (-ffriend-injection): Deprecate.
25396 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
25398         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
25399         that moved to I2, also allow destinations that are a paradoxical
25400         subreg (instead of a normal reg).
25402 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
25404         PR target/83831
25405         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
25406         to QImode.
25408 2018-02-16  Richard Biener  <rguenther@suse.de>
25410         PR tree-optimization/84037
25411         PR tree-optimization/84016
25412         PR target/82862
25413         * config/i386/i386.c (ix86_builtin_vectorization_cost):
25414         Adjust vec_construct for the fact we need additional higher latency
25415         128bit inserts for AVX256 and AVX512 vector builds.
25416         (ix86_add_stmt_cost): Scale vector construction cost for
25417         elementwise loads.
25419 2018-02-16  Richard Biener  <rguenther@suse.de>
25421         PR tree-optimization/84417
25422         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
25423         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
25424         (non_rewritable_lvalue_p): Likewise, use poly-ints.
25426 2018-02-16  Martin Liska  <mliska@suse.cz>
25428         PR sanitizer/84307
25429         * internal-fn.def (ASAN_CHECK): Set proper flags.
25430         (ASAN_MARK): Likewise.
25432 2018-02-16  Julia Koval  <julia.koval@intel.com>
25434         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
25435         from PTA_CANNONLAKE.
25437 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
25439         PR target/84272
25440         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
25441         Use ++iter rather than iter++ for std::list iterators.
25442         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
25443         defer deleting them until all nodes in the forest are processed.  Do
25444         free even leaf nodes.  Change to_process into auto_vec.
25446         PR bootstrap/84405
25447         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
25448         * vec.h (vec_default_construct): Use memset instead of placement new
25449         if BROKEN_VALUE_INITIALIZATION is defined.
25450         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
25451         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
25452         is defined.
25454         PR rtl-optimization/83723
25455         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
25456         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
25457         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
25458         recursive calls.
25459         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
25460         callers.
25461         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
25463 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
25465         PR rtl-optimization/81443
25466         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
25467         from inner REGs to paradoxical SUBREGs.
25469 2018-02-16  Richard Biener  <rguenther@suse.de>
25471         PR tree-optimization/84399
25472         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
25473         For operands we can analyze at their definition make sure we can
25474         analyze them at each use as well.
25476 2018-02-16  Richard Biener  <rguenther@suse.de>
25478         PR tree-optimization/84190
25479         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
25480         volatile accesses if the decl isn't volatile.
25482 2018-02-15  Jason Merrill  <jason@redhat.com>
25484         PR c++/84314 - ICE with templates and fastcall attribute.
25485         * attribs.c (build_type_attribute_qual_variant): Don't clobber
25486         TYPE_CANONICAL on an existing type.
25488 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
25490         PR tree-optimization/84383
25491         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
25492         dstoff nor call operand_equal_p if dstbase is NULL.
25494         PR tree-optimization/84334
25495         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
25496         also a CONSTANT_CLASS_P, punt.
25498 2018-02-14  Jim Wilson  <jimw@sifive.com>
25500         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
25501         first SMALL_OPERAND check.  New local min_second_step.  Move assert
25502         to where locals are set.  Add TARGET_RVC support.
25503         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
25505 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
25507         * doc/invoke.texi: Correct -Wformat-overflow code sample.
25509 2018-02-14  Martin Sebor  <msebor@redhat.com>
25511         PR tree-optimization/83698
25512         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
25513         arrays constrain the offset range to their bounds.
25514         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
25515         (builtin_access::overlap): Avoid setting the size of overlap if it's
25516         already been set.
25517         (maybe_diag_overlap): Also consider arrays when deciding what values
25518         of offsets to include in diagnostics.
25520 2018-02-14  Martin Sebor  <msebor@redhat.com>
25522         PR c/84108
25523         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
25524         that correspond to the kind of a declaration.
25526 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
25528         PR target/83984
25529         * config/pa/pa.md: Load address of PIC label using the linkage table
25530         if the label is nonlocal.
25532 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25534         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
25535         warning message if user requests -maltivec=be.
25536         * doc/invoke.texi: Document deprecation of -maltivec=be.
25538 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
25540         PR target/84220
25541         * config/rs6000/rs6000-c.c: Update definitions for
25542         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
25543         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
25545 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
25547         PR target/84239
25548         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
25549         add _get_ssp intrinsics. Remove argument from
25550         __builtin_ia32_rdssp[d|q].
25551         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
25552         * config/i386/i386-builtin.def: Remove argument from
25553         __builtin_ia32_rdssp[d|q].
25554         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
25555         ix86_expand_special_args_builtin for _rdssp[d|q].
25556         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
25557         Clear register before usage.
25558         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
25559         Add documentation for new _get_ssp and _inc_ssp intrinsics.
25561 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
25563         PR tree-optimization/84357
25564         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
25565         operand 1 of an ARRAY_REF too.
25567 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
25569         PR target/83831
25570         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
25571         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
25572         declarations.
25573         (set_of_reg): New struct.
25574         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
25575         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
25576         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
25577         functions.
25578         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
25579         Split into bitclr, bitset, bitinvert patterns if appropriate.
25580         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
25581         use rx_fuse_in_memory_bitop.
25582         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
25583         to named insn, correct maximum insn length.
25585 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
25587         PR target/79242
25588         * machmode.def: Define a complex mode for PARTIAL_INT.
25589         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
25590         MODE_PARTIAL_INT.
25591         * doc/rtl.texi: Document CSPImode.
25592         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
25593         handling.
25594         (msp430_hard_regno_nregs_with_padding): Likewise.
25596 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
25598         PR target/84279
25599         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
25601 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
25603         PR rtl-optimization/84169
25604         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
25605         we generated a parallel as new i3 and we split that to new i2 and i3
25606         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
25607         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
25608         those to i2, not i1.  Partially rewrite this scan code.
25610 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
25612         PR c/82210
25613         * stor-layout.c (place_field): For variable length fields, adjust
25614         offset_align afterwards not just based on the field's alignment,
25615         but also on the size.
25617         PR middle-end/84309
25618         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
25619         of exps and logs in the use_exp2 case.
25621 2018-02-13  Jeff Law  <law@redhat.com>
25623         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
25624         entry for "vector".
25626         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
25627         ARGS as unused.
25629 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
25631         PR debug/84342
25632         PR debug/84319
25633         * common.opt (gas-loc-support, gas-locview-support): New.
25634         (ginline-points, ginternal-reset-location-views): New.
25635         * doc/invoke.texi: Document them.  Use @itemx where intended.
25636         (gvariable-location-views): Adjust.
25637         * target.def (reset_location_view): New.
25638         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
25639         (TARGET_RESET_LOCATION_VIEW): New.
25640         * doc/tm.texi: Rebuilt.
25641         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
25642         (dwarf2out_default_as_locview_support): New.
25643         (output_asm_line_debug_info): Use option variables.
25644         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
25645         (output_loc_list): Likewise.
25646         (add_high_low_attributes): Check option variables.
25647         Don't output entry view attribute in strict mode.
25648         (gen_inlined_subroutine_die): Check option variables.
25649         (dwarf2out_inline_entry): Likewise.
25650         (init_sections_and_labels): Likewise.
25651         (dwarf2out_early_finish): Likewise.
25652         (maybe_reset_location_view): New, from...
25653         (dwarf2out_var_location): ... here.  Call it.
25654         * debug.h (dwarf2out_default_as_loc_support): Declare.
25655         (dwarf2out_default_as_locview_support): Declare.
25656         * hooks.c (hook_int_rtx_insn_0): New.
25657         * hooks.h (hook_int_rtx_insn_0): Declare.
25658         * toplev.c (process_options): Take -gas-loc-support and
25659         -gas-locview-support from dwarf2out.  Enable
25660         -gvariable-location-views by default only with locview
25661         assembler support.  Enable -ginternal-reset-location-views by
25662         default only if the target defines the corresponding hook.
25663         Enable -ginline-points by default if location views are
25664         enabled; force it disabled if statement frontiers are
25665         disabled.
25666         * tree-inline.c (expand_call_inline): Check option variables.
25667         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
25669 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
25671         PR tree-optimization/84321
25672         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
25673         handling.  Also check whether the anti-range contains any values
25674         that satisfy the mask; switch to a VR_RANGE if not.
25676 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
25678         PR sanitizer/84340
25679         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
25681 2018-02-13  Martin Jambor  <mjambor@suse.cz>
25683         PR c++/83990
25684         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
25685         of call statements, also set location of a load to a temporary.
25687 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
25689         * config/rl78/rl78.c (add_vector_labels): New function.
25690         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
25691         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
25692         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
25693         which checks that no arguments are passed.
25694         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
25695         * doc/extend.texi: Documentation for the new attribute.
25697 2018-02-13  Andreas Schwab  <schwab@suse.de>
25699         * config/riscv/linux.h (CPP_SPEC): Define.
25701 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
25703         PR target/84335
25704         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
25705         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
25706         OPTION_MASK_ISA_AES as first argument to def_builtin_const
25707         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
25708         instead of OPTION_MASK_ISA_PCLMUL as first argument to
25709         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
25710         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
25711         temporarily for AES and PCLMUL builtins.
25713         PR tree-optimization/84339
25714         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
25715         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
25716         Formatting fixes.
25718         PR middle-end/84309
25719         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
25720         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
25721         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
25722         inline function.
25723         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
25724         inline function.
25725         * omp-simd-clone.h: New file.
25726         * omp-simd-clone.c: Include omp-simd-clone.h.
25727         (expand_simd_clones): No longer static.
25728         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
25729         cgraph.h and omp-simd-clone.h.
25730         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
25731         (vect_recog_widen_shift_pattern): Formatting fix.
25732         (vect_pattern_recog_1): Don't check optab for calls.
25734         PR target/84336
25735         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
25736         operands[2] into a REG before using gen_lowpart on it.
25738 2018-02-12  Jeff Law  <law@redhat.com>
25740         PR target/83760
25741         * config/sh/sh.c (find_barrier): Consider a sibling call
25742         a barrier as well.
25744         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
25745         successfully back substituting a reg.
25747 2018-02-12  Richard Biener  <rguenther@suse.de>
25749         PR tree-optimization/84037
25750         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
25751         parameter, move visited init to caller.
25752         (vect_slp_analyze_operations): Separate cost from validity
25753         check, initialize visited once for all instances.
25754         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
25755         for all instances.
25756         * tree-vect-stmts.c (vect_model_simple_cost): Make early
25757         out an assert.
25758         (vect_model_promotion_demotion_cost): Likewise.
25759         (vectorizable_bswap): Guard cost modeling with !slp_node
25760         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
25761         SLP stmts.
25762         (vectorizable_call): Likewise.
25763         (vectorizable_conversion): Likewise.
25764         (vectorizable_assignment): Likewise.
25765         (vectorizable_shift): Likewise.
25766         (vectorizable_operation): Likewise.
25767         (vectorizable_store): Likewise.
25768         (vectorizable_load): Likewise.
25769         (vectorizable_condition): Likewise.
25770         (vectorizable_comparison): Likewise.
25772 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
25774         PR sanitizer/84307
25775         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
25776         (ASAN_MARK): Fix fnspec to account for return value, change pointer
25777         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
25779 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
25781         PR middle-end/83665
25782         * params.def (inline-min-speedup): Increase from 8 to 15.
25783         (max-inline-insns-auto): Decrease from 40 to 30.
25784         * ipa-split.c (consider_split): Add some buffer for function to
25785         be considered inlining candidate.
25786         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
25787         default values.
25789 2018-02-12  Richard Biener  <rguenther@suse.de>
25791         PR tree-optimization/84037
25792         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
25793         matched stmts if we cannot swap the non-matched ones.
25795 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
25797         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
25798         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
25799         _mm_maskz_scalef_round_ss): New intrinsics.
25800         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
25801         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
25802         __builtin_ia32_scalefss_round): Remove.
25803         (__builtin_ia32_scalefsd_mask_round,
25804         __builtin_ia32_scalefss_mask_round): New intrinsics.
25805         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
25806         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
25807         ((match_operand:VF_128 2 "<round_nimm_predicate>"
25808         "<round_constraint>")): Changed to ...
25809         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
25810         "<round_scalar_constraint>")): ... this.
25811         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
25812         %0, %1, %2<round_op3>}"): Changed to ...
25813         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
25814         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
25815         %2<round_scalar_mask_op3>}"): ... this.
25816         * config/i386/subst.md (round_scalar_nimm_predicate): New.
25818 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
25820         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
25821         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
25822         (_mm_maskz_sqrt_round_ss): New intrinsics.
25823         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
25824         (__builtin_ia32_sqrtsd_mask_round)
25825         (__builtin_ia32_sqrtss_mask_round): New builtins.
25826         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
25827         (__builtin_ia32_sqrtss_round): Remove.
25828         (__builtin_ia32_sqrtsd_mask_round)
25829         (__builtin_ia32_sqrtss_mask_round): New builtins.
25830         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
25831         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
25832         ((match_operand:VF_128 1 "vector_operand"
25833         "xBm,<round_constraint>")): Changed to ...
25834         ((match_operand:VF_128 1 "vector_operand"
25835         "xBm,<round_scalar_constraint>")): ... this.
25836         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
25837         %0, %2, %<iptr>1<round_op3>}): Changed to ...
25838         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
25839         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
25840         %<iptr>1<round_scalar_mask_op3>}): ... this.
25841         ((set_attr "prefix" "<round_prefix>")): Changed to ...
25842         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
25844 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
25846         PR target/84266
25847         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
25848         Cast vec_cmpeq result to correct type.
25849         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
25850         Cast vec_cmpgt result to correct type.
25852 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
25854         * final.c (final_scan_insn_1): Renamed from...
25855         (final_scan_insn): ... this.  New wrapper, to recover
25856         seen from the outermost call in recursive ones.
25857         * config/sparc/sparc.c (output_return): Drop seen from call.
25858         (output_sibcall): Likewise.
25859         * config/visium/visium.c (output_branch): Likewise.
25861 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
25863         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
25864         function label.
25866 2018-02-10  Alan Modra  <amodra@gmail.com>
25868         PR target/84300
25869         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
25870         Specify LR as an input.
25872 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
25874         PR sanitizer/83987
25875         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
25876         remove_member_access_dummy_vars): New functions.
25877         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
25878         lower_omp_1, execute_lower_omp): Use them.
25880         PR rtl-optimization/84308
25881         * shrink-wrap.c (spread_components): Release todo vector.
25883 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
25885         PR rtl-optimization/57193
25886         * ira-color.c (struct allocno_color_data): Add member
25887         conflict_allocno_hard_prefs.
25888         (update_conflict_allocno_hard_prefs): New.
25889         (bucket_allocno_compare_func): Add a preference based on
25890         conflict_allocno_hard_prefs.
25891         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
25892         (color_allocnos): Remove a dead code.  Initiate
25893         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
25895 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
25897         PR target/84226
25898         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
25899         constraint from =wa to wa.  Avoid a subreg on the output operand,
25900         instead use a pseudo and subreg it in a move.
25901         (p9_xxbrd_<mode>): Changed to ...
25902         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
25903         (p9_xxbrd_v2df): New expander.
25904         (p9_xxbrw_<mode>): Changed to ...
25905         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
25906         (p9_xxbrw_v4sf): New expander.
25908 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
25910         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
25912 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
25914         PR target/83926
25915         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
25916         multiply in 32-bit mode.
25917         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
25918         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
25919         mode.
25921 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
25923         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
25924         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
25925         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
25926         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
25928 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
25930         PR lto/84213
25931         * dwarf2out.c (is_trivial_indirect_ref): New function.
25932         (dwarf2out_late_global_decl): Do not generate a location
25933         attribute for variables that have a non-trivial DECL_VALUE_EXPR
25934         and that are not defined in the current unit.
25936 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
25938         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
25939         instead of a libcall for UNORDERED.
25941 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
25943         PR target/82641
25944         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
25945         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
25947 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
25949         PR target/PR84295
25950         * config/s390/s390.c (s390_set_current_function): Invoke
25951         s390_indirect_branch_settings also if fndecl didn't change.
25953 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
25955         * config/rs6000/rs6000.md (blockage): Set length to zero.
25957 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
25959         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
25961 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
25963         PR sanitizer/84285
25964         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
25965         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
25966         -static-lib*san.
25968         PR debug/84252
25969         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
25970         PARALLEL incoming that failed vt_get_decl_and_offset check.
25972         PR middle-end/84237
25973         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
25974         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
25975         TREE_READONLY bit.
25976         (get_variable_section): For decls in named .bss* sections pass true as
25977         second argument to bss_initializer_p.
25979 2018-02-09  Marek Polacek  <polacek@redhat.com>
25980             Jakub Jelinek  <jakub@redhat.com>
25982         PR c++/83659
25983         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
25984         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
25985         Sync some changes from cxx_fold_indirect_ref.
25987 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
25989         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
25990         markers.
25991         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
25992         (BLOCK_INLINE_ENTRY_LABEL): New.
25993         (dwarf2out_var_location): Disregard inline entry markers.
25994         (inline_entry_data): New struct.
25995         (inline_entry_data_hasher): New hashtable type.
25996         (inline_entry_data_hasher::hash): New.
25997         (inline_entry_data_hasher::equal): New.
25998         (inline_entry_data_table): New variable.
25999         (add_high_low_attributes): Add DW_AT_entry_pc and
26000         DW_AT_GNU_entry_view attributes if a pending entry is found
26001         in inline_entry_data_table.  Add old entry_pc attribute only
26002         if debug nonbinding markers are disabled.
26003         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
26004         markers are enabled.
26005         (block_within_block_p, dwarf2out_inline_entry): New.
26006         (dwarf2out_finish): Check that no entries remained in
26007         inline_entry_data_table.
26008         * final.c (reemit_insn_block_notes): Handle inline entry notes.
26009         (final_scan_insn, notice_source_line): Likewise.
26010         (rest_of_clean_state): Skip inline entry markers.
26011         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
26012         markers.
26013         * gimple.c (gimple_build_debug_inline_entry): New.
26014         * gimple.h (enum gimple_debug_subcode): Add
26015         GIMPLE_DEBUG_INLINE_ENTRY.
26016         (gimple_build_debug_inline_entry): Declare.
26017         (gimple_debug_inline_entry_p): New.
26018         (gimple_debug_nonbind_marker_p): Adjust.
26019         * insn-notes.def (INLINE_ENTRY): New.
26020         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
26021         inline entry marker notes.
26022         (print_insn): Likewise.
26023         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
26024         (INSN_DEBUG_MARKER_KIND): Likewise.
26025         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
26026         * tree-inline.c (expand_call_inline): Build and insert
26027         debug_inline_entry stmt.
26028         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
26029         inline entry blocks early, if nonbind markers are enabled.
26030         (dump_scope_block): Dump fragment info.
26031         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
26032         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
26033         (gimple_build_debug_inline_entry): New.
26034         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
26035         Enable/disable inline entry points too.
26036         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
26037         (DEBUG_INSN): Describe inline entry markers.
26039         * common.opt (gvariable-location-views): New.
26040         (gvariable-location-views=incompat5): New.
26041         * config.in: Rebuilt.
26042         * configure: Rebuilt.
26043         * configure.ac: Test assembler for view support.
26044         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
26045         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
26046         * dwarf2out.c (var_loc_view): New typedef.
26047         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
26048         (dwarf2out_locviews_in_attribute): New.
26049         (dwarf2out_locviews_in_loclist): New.
26050         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
26051         (enum dw_line_info_opcode): Add LI_adv_address.
26052         (struct dw_line_info_table): Add view.
26053         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
26054         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
26055         (zero_view_p): New variable.
26056         (ZERO_VIEW_P): New macro.
26057         (output_asm_line_debug_info): New.
26058         (struct var_loc_node): Add view.
26059         (add_AT_view_list, AT_loc_list): New.
26060         (add_var_loc_to_decl): Add view param.  Test it against last.
26061         (new_loc_list): Add view params.  Record them.
26062         (AT_loc_list_ptr): Handle loc and view lists.
26063         (view_list_to_loc_list_val_node): New.
26064         (print_dw_val): Handle dw_val_class_view_list.
26065         (size_of_die): Likewise.
26066         (value_format): Likewise.
26067         (loc_list_has_views): New.
26068         (gen_llsym): Set vl_symbol too.
26069         (maybe_gen_llsym, skip_loc_list_entry): New.
26070         (dwarf2out_maybe_output_loclist_view_pair): New.
26071         (output_loc_list): Output view list or entries too.
26072         (output_view_list_offset): New.
26073         (output_die): Handle dw_val_class_view_list.
26074         (output_dwarf_version): New.
26075         (output_compilation_unit_header): Use it.
26076         (output_skeleton_debug_sections): Likewise.
26077         (output_rnglists, output_line_info): Likewise.
26078         (output_pubnames, output_aranges): Update version comments.
26079         (output_one_line_info_table): Output view numbers in asm comments.
26080         (dw_loc_list): Determine current endview, pass it to new_loc_list.
26081         Call maybe_gen_llsym.
26082         (loc_list_from_tree_1): Adjust.
26083         (add_AT_location_description): Create view list attribute if
26084         needed, check it's absent otherwise.
26085         (convert_cfa_to_fb_loc_list): Adjust.
26086         (maybe_emit_file): Call output_asm_line_debug_info for test.
26087         (dwarf2out_var_location): Reset views as needed.  Precompute
26088         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
26089         attribute.  Set view.
26090         (new_line_info_table): Reset next view.
26091         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
26092         (dwarf2out_source_line): Likewise.  Output view resets and labels to
26093         the assembler, or select appropriate line info opcodes.
26094         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
26095         (optimize_string_length): Catch it.  Adjust.
26096         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
26097         dw_val_class_view_list, and remove it if no longer needed.
26098         (hash_loc_list): Hash view numbers.
26099         (loc_list_hasher::equal): Compare them.
26100         (optimize_location_lists): Check whether a view list symbol is
26101         needed, and whether the locview attribute is present, and
26102         whether they match.  Remove the locview attribute if no longer
26103         needed.
26104         (index_location_lists): Call skip_loc_list_entry for test.
26105         (dwarf2out_finish): Call output_asm_line_debug_info for test.
26106         Use output_dwarf_version.
26107         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
26108         (struct dw_val_node): Add val_view_list.
26109         * final.c (SEEN_NEXT_VIEW): New.
26110         (set_next_view_needed): New.
26111         (clear_next_view_needed): New.
26112         (maybe_output_next_view): New.
26113         (final_start_function): Rename to...
26114         (final_start_function_1): ... this.  Take pointer to FIRST,
26115         add SEEN parameter.  Emit param bindings in the initial view.
26116         (final_start_function): Reintroduce SEEN-less interface.
26117         (final): Rename to...
26118         (final_1): ... this.  Take SEEN parameter.  Output final pending
26119         next view at the end.
26120         (final): Reintroduce seen-less interface.
26121         (final_scan_insn): Output pending next view before switching
26122         sections or ending a block.  Mark the next view as needed when
26123         outputting variable locations.  Notify debug backend of section
26124         changes, and of location view changes.
26125         (rest_of_handle_final): Adjust.
26126         * toplev.c (process_options): Autodetect value for debug variable
26127         location views option.  Warn on incompat5 without -gdwarf-5.
26128         * doc/invoke.texi (gvariable-location-views): New.
26129         (gvariable-location-views=incompat5): New.
26130         (gno-variable-location-views): New.
26132 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
26134         PR tree-optimization/84136
26135         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
26136         that the result of find_edge is non-NULL.
26138 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
26140         PR target/83008
26141         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
26142         storing integer register in SImode.  Fix cost of 256 and 512
26143         byte aligned SSE register store.
26145 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
26147         * config/i386/i386.c (ix86_multiplication_cost): Fix
26148         multiplication cost for TARGET_AVX512DQ.
26150 2018-02-08  Marek Polacek  <polacek@redhat.com>
26152         PR tree-optimization/84238
26153         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
26154         get_range_strlen.
26156 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
26158         PR tree-optimization/84265
26159         * tree-vect-stmts.c (vectorizable_store): Don't treat
26160         VMAT_CONTIGUOUS accesses as grouped.
26161         (vectorizable_load): Likewise.
26163 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
26165         PR tree-optimization/81635
26166         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
26167         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
26168         (test_round_for_mask): New functions.
26169         (wide_int_cc_tests): Call test_round_for_mask.
26170         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
26171         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
26172         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
26173         range returned by get_range_info.
26175 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
26177         PR ipa/81360
26178         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
26179         * symtab.c: Include builtins.h
26180         (symtab_node::output_to_lto_symbol_table_p): Move here
26181         from lto-streamer-out.c:output_symbol_p.
26182         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
26183         (output_symbol_p): Move all logic to symtab.c
26184         (produce_symtab): Update.
26186 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26188         * config/s390/s390-opts.h (enum indirect_branch): Define.
26189         * config/s390/s390-protos.h (s390_return_addr_from_memory)
26190         (s390_indirect_branch_via_thunk)
26191         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
26192         (enum s390_indirect_branch_type): Define.
26193         * config/s390/s390.c (struct s390_frame_layout, struct
26194         machine_function): Remove.
26195         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
26196         (indirect_branch_table_label_no, indirect_branch_table_name):
26197         Define variables.
26198         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
26199         (enum s390_indirect_branch_option): Define.
26200         (s390_return_addr_from_memory): New function.
26201         (s390_handle_string_attribute): New function.
26202         (s390_attribute_table): Add new attribute handler.
26203         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
26204         (s390_indirect_branch_via_thunk): New function.
26205         (s390_indirect_branch_via_inline_thunk): New function.
26206         (s390_function_ok_for_sibcall): When jumping via thunk disallow
26207         sibling call optimization for non z10 compiles.
26208         (s390_emit_call): Force indirect branch target to be a single
26209         register.  Add r1 clobber for non-z10 compiles.
26210         (s390_emit_epilogue): Emit return jump via return_use expander.
26211         (s390_reorg): Handle JUMP_INSNs as execute targets.
26212         (s390_option_override_internal): Perform validity checks for the
26213         new command line options.
26214         (s390_indirect_branch_attrvalue): New function.
26215         (s390_indirect_branch_settings): New function.
26216         (s390_set_current_function): Invoke s390_indirect_branch_settings.
26217         (s390_output_indirect_thunk_function):  New function.
26218         (s390_code_end): Implement target hook.
26219         (s390_case_values_threshold): Implement target hook.
26220         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
26221         macros.
26222         * config/s390/s390.h (struct s390_frame_layout)
26223         (struct machine_function): Move here from s390.c.
26224         (TARGET_INDIRECT_BRANCH_NOBP_RET)
26225         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
26226         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
26227         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
26228         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
26229         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
26230         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
26231         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
26232         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
26233         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
26234         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
26235         (mnemonic attribute): Add values which aren't recognized
26236         automatically.
26237         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
26238         pattern for branch conversion.  Fix mnemonic attribute.
26239         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
26240         indirect branch via thunk if requested.
26241         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
26242         ("*indirect_jump"): Disable for branch conversion using out of
26243         line thunks.
26244         ("indirect_jump_via_thunk<mode>_z10")
26245         ("indirect_jump_via_thunk<mode>")
26246         ("indirect_jump_via_inlinethunk<mode>_z10")
26247         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
26248         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
26249         ("casesi_jump_via_inlinethunk<mode>_z10")
26250         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
26251         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
26252         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
26253         ("*indirect2_jump"): Disable for branch conversion.
26254         ("casesi_jump"): Turn into expander and expand patterns for branch
26255         conversion.
26256         ("return_use"): New expander.
26257         ("*return"): Emit return via thunk and rename it to ...
26258         ("*return<mode>"): ... this one.
26259         * config/s390/s390.opt: Add new options and and enum for the
26260         option values.
26262 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
26264         * lra-constraints.c (match_reload): Unconditionally use
26265         gen_lowpart_SUBREG, rather than selecting between that
26266         and equivalent gen_rtx_SUBREG code.
26268 2018-02-08  Richard Biener  <rguenther@suse.de>
26270         PR tree-optimization/84233
26271         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
26272         changed flag instead of boguously re-using phi_inserted.
26274 2018-02-08  Martin Jambor  <mjambor@suse.cz>
26276         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
26277         static local variables.
26279 2018-02-08  Richard Biener  <rguenther@suse.de>
26281         PR tree-optimization/84278
26282         * tree-vect-stmts.c (vectorizable_store): When looking for
26283         smaller vector types to perform grouped strided loads/stores
26284         make sure the mode is supported by the target.
26285         (vectorizable_load): Likewise.
26287 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
26289         * config/aarch64/aarch64.c (aarch64_components_for_bb):
26290         Increase LDP/STP opportunities by adding adjacent callee-saves.
26292 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
26294         PR rtl-optimization/84068
26295         PR rtl-optimization/83459
26296         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
26298 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
26300         PR tree-optimization/84224
26301         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
26302         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
26303         non-zero arguments.
26305 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
26307         PR target/84113
26308         * config/rs6000/altivec.md (*restore_world): Remove LR use.
26309         * config/rs6000/predicates.md (restore_world_operation): Adjust op
26310         count, remove one USE.
26312 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
26314         * doc/install.texi (Configuration): Document the
26315         --with-long-double-format={ibm,ieee} PowerPC configuration
26316         options.
26318         PR target/84154
26319         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
26320         Convert from define_expand to be define_insn_and_split.  Rework
26321         float/double/_Float128 conversions to QI/HI/SImode to work with
26322         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
26323         conversions to QI/HImode types did a store and then a load to
26324         truncate the value.  For conversions to VSX registers, don't split
26325         the insn, instead emit the code directly.  Use the code iterator
26326         any_fix to combine signed and unsigned conversions.
26327         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
26328         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
26329         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
26330         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
26331         (fix_<mode>di2_hw): Likewise.
26332         (fixuns_<mode>di2_hw): Likewise.
26333         (fix_<mode>si2_hw): Likewise.
26334         (fixuns_<mode>si2_hw): Likewise.
26335         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
26336         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
26337         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
26338         fix<uns>_trunc<SFDF:mode>si2_p8.
26339         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
26340         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
26341         (fix<uns>_<mode>_mem): Likewise.
26342         (fctiw<u>z_<mode>_mem): Likewise.
26343         (fix<uns>_<mode>_mem): Likewise.
26344         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
26345         the register allocator from doing a direct move to the GPRs to do
26346         a store, and instead use the ISA 3.0 store byte/half-word from
26347         vector register instruction.  For IEEE 128-bit floating point,
26348         also optimize stores of 32-bit ints.
26349         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
26351 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
26353         * genextract.c (push_pathstr_operand): New function to support
26354         [a-zA-Z].
26355         (walk_rtx): Call push_pathstr_operand.
26356         (print_path): Support [a-zA-Z].
26358 2018-02-07  Richard Biener  <rguenther@suse.de>
26360         PR tree-optimization/84037
26361         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
26362         (cse_and_gimplify_to_preheader): Declare.
26363         (vect_get_place_in_interleaving_chain): Likewise.
26364         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
26365         ivexpr_map.
26366         (_loop_vec_info::~_loop_vec_info): Delete it.
26367         (cse_and_gimplify_to_preheader): New function.
26368         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
26369         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
26370         (vectorizable_load): Likewise.  For grouped stores always base
26371         the IV on the first element.
26372         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
26373         condition before gimplifying.
26375 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
26377         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
26378         *DIV_EXPR and *MOD_EXPR.
26380 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
26382         PR target/84248
26383         * config/i386/i386.c (ix86_option_override_internal): Mask out
26384         the CF_SET bit when checking -fcf-protection.
26386 2018-02-07  Tom de Vries  <tom@codesourcery.com>
26388         PR libgomp/84217
26389         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
26390         enough.
26392 2018-02-07  Richard Biener  <rguenther@suse.de>
26394         PR tree-optimization/84204
26395         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
26396         this place.
26398         PR tree-optimization/84205
26399         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
26400         special-case isl_ast_op_zdiv_r.
26402         PR tree-optimization/84223
26403         * graphite-scop-detection.c (gather_bbs::before_dom_children):
26404         Only add conditions from within the region.
26405         (gather_bbs::after_dom_children): Adjust.
26407 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
26409         PR target/84209
26410         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
26411         * config/avr/avr.md: Only post-reload split REG-REG moves if
26412         either register is GENERAL_REG_P.
26414 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
26416         PR tree-optimization/84235
26417         * tree-ssa-scopedtables.c
26418         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
26419         if the subtraction is performed in floating point type where NaNs are
26420         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
26421         build 1.  Formatting fix.
26423 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
26425         PR target/84146
26426         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
26427         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
26428         and skip it regardless of bb boundaries.  Use CALL_P macro,
26429         don't test INSN_P (insn) together with CALL_P or JUMP_P check
26430         unnecessarily, formatting fix.
26432 2018-02-06  Michael Collison  <michael.collison@arm.com>
26434         * config/arm/thumb2.md:
26435         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
26436         (*thumb_mov_notscc): Ditto.
26438 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
26440         PR target/84154
26441         * config/rs6000/rs6000.md (su code attribute): Use "u" for
26442         unsigned_fix, not "s".
26444 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26446         * configure.ac (gcc_fn_eh_frame_ro): New function.
26447         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
26448         correct .eh_frame permissions.
26449         * configure: Regenerate.
26451 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
26453         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
26454         irrelevant options.
26456 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26458         * config/rs6000/rs6000.c (rs6000_option_override_internal):
26459         Display warning message for -mno-speculate-indirect-jumps.
26461 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
26463         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
26464         Undocumented.
26465         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
26467 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
26469         PR tree-optimization/84225
26470         * tree-eh.c (find_trapping_overflow): Only call
26471         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
26473 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
26475         PR target/84145
26476         * config/i386/i386.c: Reimplement the check of possible options
26477         -mibt/-mshstk conbination. Change error messages.
26478         * doc/invoke.texi: Fix a typo: remove extra '='.
26480 2018-02-06  Marek Polacek  <polacek@redhat.com>
26482         PR tree-optimization/84228
26483         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
26485 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
26487         PR target/82641
26488         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
26489         emitted arch directives.
26490         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
26491         __ARM_FEATURE_COPROC before changing architectures.
26493 2018-02-06  Richard Biener  <rguenther@suse.de>
26495         * config/i386/i386.c (print_reg): Fix typo.
26496         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
26498 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
26500         * configure: Regenerate.
26502 2018-02-05  Martin Sebor  <msebor@redhat.com>
26504         PR tree-optimization/83369
26505         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
26506         inlining context.
26508 2018-02-05  Martin Liska  <mliska@suse.cz>
26510         * doc/invoke.texi: Cherry-pick upstream r323995.
26512 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
26514         * ira.c (ira_init_register_move_cost): Adjust comment.
26516 2018-02-05  Martin Liska  <mliska@suse.cz>
26518         PR gcov-profile/84137
26519         * doc/gcov.texi: Fix typo in documentation.
26521 2018-02-05  Martin Liska  <mliska@suse.cz>
26523         PR gcov-profile/83879
26524         * doc/gcov.texi: Document necessity of --dynamic-list-data when
26525         using dlopen functionality.
26527 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
26529         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
26530         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
26531         _mm_maskz_range_ss, _mm_mask_range_round_ss,
26532         _mm_maskz_range_round_ss): New intrinsics.
26533         (__builtin_ia32_rangesd128_round)
26534         (__builtin_ia32_rangess128_round): Remove.
26535         (__builtin_ia32_rangesd128_mask_round,
26536         __builtin_ia32_rangess128_mask_round): New builtins.
26537         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
26538         __builtin_ia32_rangess128_round): Remove.
26539         (__builtin_ia32_rangesd128_mask_round,
26540         __builtin_ia32_rangess128_mask_round): New builtins.
26541         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
26542         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
26543         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
26544         "<round_saeonly_constraint>")): Changed to ...
26545         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
26546         "<round_saeonly_scalar_constraint>")): ... this.
26547         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
26548         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
26549         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
26550         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
26551         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
26553 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
26555         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
26556         options.
26557         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
26558         Remove all values except native, 8540 and 8548.
26560 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
26562         * config/i386/i386.c (ix86_output_function_return): Pass
26563         INVALID_REGNUM, instead of -1, as invalid register number to
26564         indirect_thunk_name and output_indirect_thunk.
26566 2018-02-02  Julia Koval  <julia.koval@intel.com>
26568         * config.gcc: Add -march=icelake.
26569         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
26570         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
26571         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
26572         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
26573         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
26574         (processor_target_table): Add icelake.
26575         (ix86_option_override_internal): Handle new PTAs.
26576         (get_builtin_code_for_version): Handle icelake.
26577         (M_INTEL_COREI7_ICELAKE): New.
26578         (fold_builtin_cpu): Handle icelake.
26579         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
26580         * doc/invoke.texi: Add -march=icelake.
26582 2018-02-02  Julia Koval  <julia.koval@intel.com>
26584         * config/i386/i386.c (ix86_option_override_internal): Change flags type
26585         to wide_int_bitmask.
26586         * wide-int-bitmask.h: New.
26588 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
26590         PR target/84066
26591         * config/i386/i386.md: Replace Pmode with word_mode in
26592         builtin_setjmp_setup and builtin_longjmp to support x32.
26594 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
26596         PR target/56010
26597         PR target/83743
26598         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
26599         #include "opts.h".
26600         (rs6000_supported_cpu_names): New static variable.
26601         (linux_cpu_translation_table): Likewise.
26602         (elf_platform) <cpu>: Define new static variable and use it.
26603         Translate kernel AT_PLATFORM name to canonical name if needed.
26604         Error if platform name is unknown.
26606 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
26608         PR target/84089
26609         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
26611 2018-02-01  Jeff Law  <law@redhat.com>
26613         PR target/84128
26614         * config/i386/i386.c (release_scratch_register_on_entry): Add new
26615         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
26616         the scratch if RELEASE_VIA_POP is false.
26617         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
26618         If we have to save a temporary register, decrement SIZE appropriately.
26619         Pass new arguments to release_scratch_register_on_entry.
26620         (ix86_adjust_stack_and_probe): Likewise.
26621         (ix86_emit_probe_stack_range): Pass new arguments to
26622         release_scratch_register_on_entry.
26624 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
26626         PR rtl-optimization/84157
26627         * combine.c (change_zero_ext): Use REG_P predicate in
26628         front of HARD_REGISTER_P predicate.
26630 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
26632         * config/avr/avr.c (avr_option_override): Move disabling of
26633         -fdelete-null-pointer-checks to...
26634         * common/config/avr/avr-common.c (avr_option_optimization_table):
26635         ...here.
26637 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26639         PR tree-optimization/81635
26640         * tree-data-ref.c (split_constant_offset_1): For types that
26641         wrap on overflow, try to use range info to prove that wrapping
26642         cannot occur.
26644 2018-02-01  Renlin Li  <renlin.li@arm.com>
26646         PR target/83370
26647         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
26648         TAILCALL_ADDR_REGS.
26649         (aarch64_register_move_cost): Likewise.
26650         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
26651         TAILCALL_ADDR_REGS.
26652         (REG_CLASS_NAMES): Likewise.
26653         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
26654         TAILCALL_ADDR_REGS. Remove IP registers.
26655         * config/aarch64/aarch64.md (Ucs): Update register constraint.
26657 2018-02-01  Richard Biener  <rguenther@suse.de>
26659         * domwalk.h (dom_walker::dom_walker): Add additional constructor
26660         for specifying RPO order and allow NULL for that.
26661         * domwalk.c (dom_walker::dom_walker): Likewise.
26662         (dom_walker::walk): Handle NULL RPO order.
26663         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
26664         in RPO order.
26665         (rewrite_update_dom_walker): Likewise.
26666         (mark_def_dom_walker): Likewise.
26668 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26670         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
26671         (aarch64_maybe_expand_sve_subreg_move): Declare.
26672         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
26673         * config/aarch64/predicates.md (aarch64_any_register_operand): New
26674         predicate.
26675         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
26676         that are semantically a reverse operation.
26677         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
26678         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
26679         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
26680         functions.
26681         (aarch64_can_change_mode_class): For big-endian, forbid changes
26682         between two SVE modes if they have different element sizes.
26684 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26686         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
26687         the TImode handling for big-endian targets.
26689 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26691         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
26692         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
26693         not just bytes.
26694         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
26695         Remove BSWAP handing for big-endian targets and use the form of
26696         LD1RQ appropariate for the mode.
26698 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26700         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
26701         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
26702         duplicated element.
26704 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26706         PR tearget/83845
26707         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
26708         check for operands that need to go through aarch64_sve_reload_be.
26710 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
26712         PR tree-optimization/81661
26713         PR tree-optimization/84117
26714         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
26715         * tree-eh.c: Include gimplify.h.
26716         (find_trapping_overflow, replace_trapping_overflow,
26717         rewrite_to_non_trapping_overflow): New functions.
26718         * tree-vect-loop.c: Include tree-eh.h.
26719         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
26720         * tree-data-ref.c: Include tree-eh.h.
26721         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
26723 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
26725         PR rtl-optimization/84123
26726         * combine.c (change_zero_ext): Check if hard register satisfies
26727         can_change_dest_mode before calling gen_lowpart_SUBREG.
26729 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
26731         PR target/82444
26732         * ira.c (ira_init_register_move_cost): Remove assert.
26734 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26736         PR rtl-optimization/84071
26737         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
26738         * doc/tm.texi: Regenerate.
26740 2018-01-31  Richard Biener  <rguenther@suse.de>
26742         PR tree-optimization/84132
26743         * tree-data-ref.c (analyze_miv_subscript): Properly
26744         check whether evolution_function_is_affine_multivariate_p
26745         before calling gcd_of_steps_may_divide_p.
26747 2018-01-31  Julia Koval  <julia.koval@intel.com>
26749         PR target/83618
26750         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
26751         * config/i386/i386.md (rdpid_rex64) New.
26752         (rdpid): Make 32bit only.
26754 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
26756         PR lto/84105
26757         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
26758         an IDENTIFIER_NODE for FUNCTION_TYPE's.
26760 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26762         Revert
26763         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
26765         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
26767 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26769         PR rtl-optimization/84071
26770         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
26771         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
26773 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
26775         * config/arc/arc.c (arc_handle_aux_attribute): New function.
26776         (arc_attribute_table): Add 'aux' attribute.
26777         (arc_in_small_data_p): Consider aux like variables.
26778         (arc_is_aux_reg_p): New function.
26779         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
26780         (arc_get_aux_arg): New function.
26781         (prepare_move_operands): Handle aux-register access.
26782         (arc_handle_aux_attribute): New function.
26783         * doc/extend.texi (ARC Variable attributes): Add subsection.
26785 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
26787         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
26788         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
26789         (arc_attribute_table): Add 'uncached' attribute.
26790         (arc_print_operand): Print '.di' flag for uncached memory
26791         accesses.
26792         (arc_in_small_data_p): Do not consider for small data the uncached
26793         types.
26794         (arc_is_uncached_mem_p): New function.
26795         * config/arc/predicates.md (compact_store_memory_operand): Check
26796         for uncached memory accesses.
26797         (nonvol_nonimm_operand): Likewise.
26798         * doc/extend.texi (ARC Type Attribute): New subsection.
26800 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
26802         PR c/84100
26803         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
26804         falign-loops=): Add Optimization flag.
26806 2018-01-30  Jeff Law  <law@redhat.com>
26808         PR target/84064
26809         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
26810         INT_REGISTERS_SAVED.  Check it prior to calling
26811         get_scratch_register_on_entry.
26812         (ix86_adjust_stack_and_probe): Similarly.
26813         (ix86_emit_probe_stack_range): Similarly.
26814         (ix86_expand_prologue): Corresponding changes.
26816 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26818         PR target/40411
26819         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
26820         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
26822 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
26824         PR target/84112
26825         * lra-constraints.c (curr_insn_transform): Process AND in the
26826         address.
26828 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
26830         PR rtl-optimization/83986
26831         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
26832         dependence against last_pending_memory_flush in addition to
26833         pending_jump_insns.
26835 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
26837         PR tree-optimization/81611
26838         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
26839         copies.
26841 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26843         PR target/83758
26844         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
26845         a reg rtx.
26847 2018-01-30  Richard Biener  <rguenther@suse.de>
26848             Jakub Jelinek  <jakub@redhat.com>
26850         PR tree-optimization/84111
26851         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
26852         inner loops added during recursion, as they don't have up-to-date
26853         SSA form.
26855 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
26857         PR ipa/81360
26858         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
26859         (can_inline_edge_by_limits_p): ... here.
26860         (can_early_inline_edge_p, check_callers,
26861         update_caller_keys, update_callee_keys, recursive_inlining,
26862         add_new_edges_to_heap, speculation_useful_p,
26863         inline_small_functions,
26864         inline_small_functions, flatten_function,
26865         inline_to_all_callers_1): Update.
26867 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
26869         * profile-count.c (profile_count::combine_with_ipa_count): Handle
26870         zeros correctly.
26872 2018-01-30  Richard Biener  <rguenther@suse.de>
26874         PR tree-optimization/83008
26875         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
26876         invariant and constant vector uses in stmts when they need
26877         more than one stmt.
26879 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26881         PR bootstrap/84017
26882         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
26883         * configure: Regenerate.
26885 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26887         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
26888         pattern.
26889         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
26890         Use gen_rtx_REG rather than gen_lowpart.
26892 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26894         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
26895         rather than 0 when creating partial subregs.
26897 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26899         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
26900         of usage.
26902 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
26904         PR target/81550
26905         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
26906         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
26907         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
26908         flags.  This restores the settings used before the 2017-07-24.
26909         Turning off pre increment/decrement/modify allows IVOPTS to
26910         optimize DF/SF loops where the index is an int.
26912 2018-01-29  Richard Biener  <rguenther@suse.de>
26913             Kelvin Nilsen  <kelvin@gcc.gnu.org>
26915         PR bootstrap/80867
26916         * tree-vect-stmts.c (vectorizable_call): Don't call
26917         targetm.vectorize_builtin_md_vectorized_function if callee is
26918         NULL.
26920 2018-01-22  Carl Love  <cel@us.ibm.com>
26922         * doc/extend.tex: Fix typo in second arg in
26923         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
26925 2018-01-29  Richard Biener  <rguenther@suse.de>
26927         PR tree-optimization/84086
26928         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
26929         (flush_ssaname_freelist): When SSA names were released reset
26930         the SCEV hash table.
26932 2018-01-29  Richard Biener  <rguenther@suse.de>
26934         PR tree-optimization/84057
26935         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
26936         removed paths when removing edges.
26938 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
26940         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
26941         -mfunction-return=@var{choice}.
26943 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
26945         PR diagnostic/84034
26946         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
26947         Handle CR like TAB.
26948         (layout::print_source_line): Likewise.
26949         (test_get_line_width_without_trailing_whitespace): Add test cases.
26951 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
26953         PR middle-end/84040
26954         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
26955         debug insns.
26957 2018-01-26  Jim Wilson  <jimw@sifive.com>
26959         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
26961         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
26962         specified.
26964 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26966         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
26967         and CMP + SUB-immediate -> SUBS.
26969 2018-01-26  Martin Sebor  <msebor@redhat.com>
26971         PR tree-optimization/83896
26972         * tree-ssa-strlen.c (get_string_len): Rename...
26973         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
26974         Avoid assuming length is constant.
26975         (handle_char_store): Use HOST_WIDE_INT for string length.
26977 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
26979         PR target/81763
26980         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
26981         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
26983 2018-01-26  Richard Biener  <rguenther@suse.de>
26985         PR rtl-optimization/84003
26986         * dse.c (record_store): Only record redundant stores when
26987         the earlier store aliases at least all accesses the later one does.
26989 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
26991         PR rtl-optimization/83985
26992         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
26993         REG_CFA_RESTORE insns.
26994         (delete_unmarked_insns): Don't ignore separate shrink wrapping
26995         REG_CFA_RESTORE insns here.
26997         PR c/83989
26998         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
26999         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
27001 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27003         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
27004         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
27005         (arc_init): Likewise.
27006         (arc_override_options): Likewise.
27007         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
27008         value.
27009         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
27010         support.
27011         * config/arc/arc.h (TARGET_DBNZ): Define.
27012         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
27013         properly set the tune attribute.
27014         (dbnz): Use TARGET_DBNZ guard.
27015         * config/arc/arc.opt (mtune): Add core3 option.
27017 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27019         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
27020         recognize new pic like addresses.
27021         (arc_delegitimize_address): Clean up.
27023 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27025         * config/arc/arc-arches.def: Option mrf16 valid for all
27026         architectures.
27027         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
27028         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
27029         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
27030         * config/arc/arc-tables.opt: Regenerate.
27031         * config/arc/arc.c (arc_conditional_register_usage): Handle
27032         reduced register file case.
27033         (arc_file_start): Set must have build attributes.
27034         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
27035         mrf16 option value.
27036         * config/arc/arc.opt (mrf16): Add new option.
27037         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
27038         * config/arc/genmultilib.awk: Handle new mrf16 option.
27039         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
27040         * config/arc/t-multilib: Regenerate.
27041         * doc/invoke.texi (ARC Options): Document mrf16 option.
27043 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27045         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
27046         * config/arc/arc.c (arc_handle_secure_attribute): New function.
27047         (arc_attribute_table): Add 'secure_call' attribute.
27048         (arc_print_operand): Print secure call operand.
27049         (arc_function_ok_for_sibcall): Don't optimize tail calls when
27050         secure.
27051         (arc_is_secure_call_p): New function.  * config/arc/arc.md
27052         (call_i): Add support for sjli instruction.
27053         (call_value_i): Likewise.
27054         * config/arc/constraints.md (Csc): New constraint.
27056 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27057             John Eric Martin  <John.Martin@emmicro-us.com>
27059         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
27060         * config/arc/arc.c (_arc_jli_section): New struct.
27061         (arc_jli_section): New type.
27062         (rc_jli_sections): New static variable.
27063         (arc_handle_jli_attribute): New function.
27064         (arc_attribute_table): Add jli_always and jli_fixed attribute.
27065         (arc_file_end): New function.
27066         (TARGET_ASM_FILE_END): Define.
27067         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
27068         (arc_add_jli_section): New function.
27069         (jli_call_scan): Likewise.
27070         (arc_reorg): Call jli_call_scan.
27071         (arc_output_addsi): Remove 'S' from printing asm operand.
27072         (arc_is_jli_call_p): New function.
27073         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
27074         operand.
27075         (movhi_insn): Likewise.
27076         (movsi_insn): Likewise.
27077         (movsi_set_cc_insn): Likewise.
27078         (loadqi_update): Likewise.
27079         (load_zeroextendqisi_update): Likewise.
27080         (load_signextendqisi_update): Likewise.
27081         (loadhi_update): Likewise.
27082         (load_zeroextendhisi_update): Likewise.
27083         (load_signextendhisi_update): Likewise.
27084         (loadsi_update): Likewise.
27085         (loadsf_update): Likewise.
27086         (movsicc_insn): Likewise.
27087         (bset_insn): Likewise.
27088         (bxor_insn): Likewise.
27089         (bclr_insn): Likewise.
27090         (bmsk_insn): Likewise.
27091         (bicsi3_insn): Likewise.
27092         (cmpsi_cc_c_insn): Likewise.
27093         (movsi_ne): Likewise.
27094         (movsi_cond_exec): Likewise.
27095         (clrsbsi2): Likewise.
27096         (norm_f): Likewise.
27097         (normw): Likewise.
27098         (swap): Likewise.
27099         (divaw): Likewise.
27100         (flag): Likewise.
27101         (sr): Likewise.
27102         (kflag): Likewise.
27103         (ffs): Likewise.
27104         (ffs_f): Likewise.
27105         (fls): Likewise.
27106         (call_i): Remove 'S' asm letter, add jli instruction.
27107         (call_value_i): Likewise.
27108         * config/arc/arc.op (mjli-always): New option.
27109         * config/arc/constraints.md (Cji): New constraint.
27110         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
27111         operand.
27112         (subsf3_fpx): Likewise.
27113         (mulsf3_fpx): Likewise.
27114         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
27115         asm operand.
27116         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
27117         function attrbutes.
27118         * doc/invoke.texi (ARC): Document mjli-always option.
27120 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
27122         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
27123         avoid addition with 0 and use incw and decw where possible.
27125 2018-01-26  Richard Biener  <rguenther@suse.de>
27127         PR tree-optimization/81082
27128         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
27129         association if it requires casting to unsigned.
27130         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
27131         from fold_plusminus_mult_expr to catch important cases late when
27132         range info is available.
27134 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27136         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
27137         * configure.ac (hidden_linkonce): New test.
27138         * configure: Regenerate.
27139         * config.in: Regenerate.
27141 2018-01-26  Julia Koval  <julia.koval@intel.com>
27143         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
27144         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
27145         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
27146         _mm_mask_bitshuffle_epi64_mask): Fix type.
27147         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
27148         USI_FTYPE_V4DI_V4DI_USI): Remove.
27149         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
27150         __builtin_ia32_vpshufbitqmb256_mask,
27151         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
27152         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
27153         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
27155 2018-01-26  Alan Modra  <amodra@gmail.com>
27157         PR target/84033
27158         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
27159         UNSPEC_VBPERMQ.  Sort other unspecs.
27161 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
27163         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
27165 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
27167         PR middle-end/83055
27168         * predict.c (drop_profile): Do not push/pop cfun; update also
27169         node->count.
27170         (handle_missing_profiles): Fix logic looking for zero profiles.
27172 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
27174         PR middle-end/83977
27175         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
27176         on functions with #pragma omp declare simd or functions with simd
27177         attribute.
27178         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
27179         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
27180         Remove trailing \n from warning_at calls.
27182 2018-01-25  Tom de Vries  <tom@codesourcery.com>
27184         PR target/84028
27185         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
27186         for neutered workers.
27188 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
27190         PR target/68467
27191         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
27192         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
27194 2018-01-24  Jeff Law  <law@redhat.com>
27196         PR target/83994
27197         * i386.c (get_probe_interval): Move to earlier point.
27198         (ix86_compute_frame_layout): If -fstack-clash-protection and
27199         the frame is larger than the probe interval, then use pushes
27200         to save registers rather than reg->mem moves.
27201         (ix86_expand_prologue): Remove conditional for int_registers_saved
27202         assertion.
27204 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
27206         PR target/84014
27207         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
27208         min/max for never referenced object.
27210 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
27212         PR middle-end/83977
27213         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
27214         here.
27215         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
27216         attributes from DECL_ATTRIBUTES (decl) without affecting
27217         DECL_ATTRIBUTES (current_function_decl).
27218         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
27219         functions with non-NULL DECL_ABSTRACT_ORIGIN.
27221 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
27223         PR tree-optimization/83979
27224         * fold-const.c (fold_comparison): Use constant_boolean_node
27225         instead of boolean_{true,false}_node.
27227 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
27229         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
27230         with zero counts.
27232 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27234         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27235         Simplify the clause that sets the length attribute.
27236         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27237         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
27238         clause that sets the length attribute.
27239         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27241 2018-01-24  Tom de Vries  <tom@codesourcery.com>
27243         PR target/83589
27244         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
27245         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
27246         Add strict parameter.
27247         (prevent_branch_around_nothing): Insert dummy insn between branch to
27248         label and label with no ptx insn inbetween.
27249         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
27251 2018-01-24  Tom de Vries  <tom@codesourcery.com>
27253         PR target/81352
27254         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
27255         for neutered threads in warp.
27256         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
27258 2018-01-24  Richard Biener  <rguenther@suse.de>
27260         PR tree-optimization/83176
27261         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
27262         operands.
27264 2018-01-24  Richard Biener  <rguenther@suse.de>
27266         PR tree-optimization/82819
27267         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
27268         code generating pluses that are no-ops in the target precision.
27270 2018-01-24  Richard Biener  <rguenther@suse.de>
27272         PR middle-end/84000
27273         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
27275 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
27277         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
27278         to merge probabilities.
27279         * predict.c (probably_never_executed): Also mark as cold functions
27280         with global 0 profile and guessed local profile.
27281         * profile-count.c (profile_probability::combine_with_count): New
27282         member function.
27283         * profile-count.h (profile_probability::operator*,
27284         profile_probability::operator*=, profile_probability::operator/,
27285         profile_probability::operator/=): Reduce precision to adjusted
27286         and set value to guessed on contradictory divisions.
27287         (profile_probability::combine_with_freq): Remove.
27288         (profile_probability::combine_wiht_count): Declare.
27289         (profile_count::force_nonzero):: Set to adjusted.
27290         (profile_count::probability_in):: Set quality to adjusted.
27291         * tree-ssa-tail-merge.c (replace_block_by): Use
27292         combine_with_count.
27294 2018-01-23  Andrew Waterman  <andrew@sifive.com>
27295             Jim Wilson  <jimw@sifive.com>
27297         * config/riscv/riscv.c (riscv_stack_boundary): New.
27298         (riscv_option_override): Set riscv_stack_boundary.  Handle
27299         riscv_preferred_stack_boundary_arg.
27300         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
27301         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
27302         (STACK_BOUNDARY): Set to riscv_stack_boundary.
27303         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
27304         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
27305         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
27307 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
27309         PR target/83905
27310         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
27311         of struct ix86_frame.
27312         (ix86_expand_epilogue): Likewise.  Add a local variable for
27313         the reg_save_offset field in struct ix86_frame.
27315 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
27317         PR tree-optimization/82604
27318         * tree-loop-distribution.c (enum partition_kind): New enum item
27319         PKIND_PARTIAL_MEMSET.
27320         (partition_builtin_p): Support above new enum item.
27321         (generate_code_for_partition): Ditto.
27322         (compute_access_range): Differentiate cases that equality can be
27323         proven at all loops, the innermost loops or no loops.
27324         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
27325         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
27326         (finalize_partitions, distribute_loop): Don't fuse partition of
27327         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
27328         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
27329         parloop is enabled.
27331 2018-01-23  Martin Liska  <mliska@suse.cz>
27333         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
27334         order to ignore the predictor.
27335         (PRED_POLYMORPHIC_CALL): Likewise.
27336         (PRED_RECURSIVE_CALL): Likewise.
27338 2018-01-23  Martin Liska  <mliska@suse.cz>
27340         * tree-profile.c (tree_profiling): Print function header to
27341         aware reader which function we are working on.
27342         * value-prof.c (gimple_find_values_to_profile): Do not print
27343         not interesting value histograms.
27345 2018-01-23  Martin Liska  <mliska@suse.cz>
27347         * profile-count.h (enum profile_quality): Add
27348         profile_uninitialized as the first value. Do not number values
27349         as they are zero based.
27350         (profile_count::verify): Update sanity check.
27351         (profile_probability::verify): Likewise.
27353 2018-01-23  Nathan Sidwell  <nathan@acm.org>
27355         * doc/invoke.texi (ffor-scope): Deprecate.
27357 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
27359         PR tree-optimization/83510
27360         * domwalk.c (set_all_edges_as_executable): New function.
27361         (dom_walker::dom_walker): Convert bool param
27362         "skip_unreachable_blocks" to enum reachability.  Move setup of
27363         edge flags to set_all_edges_as_executable and only do it when
27364         reachability is REACHABLE_BLOCKS.
27365         * domwalk.h (enum dom_walker::reachability): New enum.
27366         (dom_walker::dom_walker): Convert bool param
27367         "skip_unreachable_blocks" to enum reachability.
27368         (set_all_edges_as_executable): New decl.
27369         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
27370         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
27371         "reachability".
27372         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
27373         but converting true to REACHABLE_BLOCKS.
27374         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
27375         * tree-vrp.c
27376         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
27377         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
27378         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
27379         REACHABLE_BLOCKS.
27380         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
27381         if check_all_array_refs will be called.
27383 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
27385         * tree.c (selftest::test_location_wrappers): Add more test
27386         coverage.
27388 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
27390         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
27391         (selftest::test_bit_in_range): Likewise.
27393 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
27395         PR testsuite/83888
27396         * doc/sourcebuild.texi (vect_float): Say that the selector
27397         only describes the situation when -funsafe-math-optimizations is on.
27398         (vect_float_strict): Document.
27400 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
27402         PR tree-optimization/83965
27403         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
27404         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
27405         instead of checking only for a reduction.
27406         (vect_recog_widen_sum_pattern): Likewise.
27408 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
27410         * predict.c (probably_never_executed): Only use precise profile info.
27411         (compute_function_frequency): Skip after inlining hack since we now
27412         have quality checking.
27414 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
27416         * profile-count.h (profile_probability::very_unlikely,
27417         profile_probability::unlikely, profile_probability::even): Set
27418         precision to guessed.
27420 2018-01-23  Richard Biener  <rguenther@suse.de>
27422         PR tree-optimization/83963
27423         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
27424         Properly terminate dominator walk when crossing the exit edge not
27425         when visiting its source block.
27427 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
27429         PR c++/83918
27430         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
27431         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
27433 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
27435         PR tree-optimization/83957
27436         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
27437         semicolon after for body surrounded by braces.
27439         PR tree-optimization/83081
27440         * profile-count.h (profile_probability::split): New method.
27441         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
27442         Use profile_probability::split.
27443         (do_compare_rtx_and_jump): Fix adjustment of probabilities
27444         when splitting a single conditional jump into 2.
27446 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
27448         PR tree-optimization/69452
27449         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
27450         decl.
27452 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27454         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
27455         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
27456         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
27458 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27460         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
27461         declaration.
27462         * config/rl78/rl78.md (movdi): New define_expand.
27463         * config/rl78/rl78.c (rl78_split_movdi): New function.
27465 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
27467         PR target/83862
27468         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
27469         no longer used.
27470         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
27471         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
27472         128-bit to produce an UNSPEC move to get the double word with the
27473         signbit and then a shift directly to do signbit.
27474         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
27475         implementation with a new version that just does either a direct
27476         move or a regular move.  Move memory interface to separate insns.
27477         Move insns so they are next to the expander.
27478         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
27479         with signbit move.  Split big and little endian case.
27480         (signbit<mode>2_dm_mem_le): Likewise.
27481         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
27482         (signbit<mode>2_dm2): Likewise.
27484 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27486         * config/rl78/rl78.md (anddi3): New define_expand.
27488 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27490         * config/rl78/rl78.md (umindi3): New define_expand.
27492 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27494         * config/rl78/rl78.md (smindi3): New define_expand.
27496 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27498         * config/rl78/rl78.md (smaxdi3): New define_expand.
27500 2018-01-22  Carl Love  <cel@us.ibm.com>
27502         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
27503         LVX_V1TI): Add macro expansion.
27504         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
27505         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
27506         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
27507         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
27508         Change check to determine if the instruction is a byte reversing
27509         entry.  Fix typo in comment.
27510         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
27511         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
27512         Add def_builtin calls for new builtins.
27513         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
27514         Add define_insn expansion.
27516 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27518         * config/rl78/rl78.md (umaxdi3): New define_expand.
27520 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27522         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
27523         for non-QImode registers.
27525 2018-01-22  Richard Biener  <rguenther@suse.de>
27527         PR tree-optimization/83963
27528         * graphite-scop-detection.c (scop_detection::get_sese): Delay
27529         including the loop exit block.
27530         (scop_detection::merge_sese): Likewise.
27531         (scop_detection::add_scop): Do it here instead.
27533 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27535         * doc/sourcebuild.texi (arm_softfloat): Document.
27537 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
27539         PR gcc/77734
27540         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
27541         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
27542         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
27544 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27545             David Edelsohn  <dje.gcc@gmail.com>
27547         PR target/83946
27548         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27549         Change "crset eq" to "crset 2".
27550         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27551         (*call_indirect_aix<mode>_nospec): Likewise.
27552         (*call_value_indirect_aix<mode>_nospec): Likewise.
27553         (*call_indirect_elfv2<mode>_nospec): Likewise.
27554         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
27555         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
27556         change assembly output from . to $.
27557         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27558         (indirect_jump<mode>_nospec): Change assembly output from . to $.
27559         (*tablejump<mode>_internal1_nospec): Likewise.
27561 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
27563         PR target/80870
27564         * config/sh/sh_optimize_sett_clrt.cc:
27565         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
27567 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
27569         PR tree-optimization/83940
27570         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
27571         offset_dt to vect_constant_def rather than vect_unknown_def_type.
27572         (vect_check_load_store_mask): Add a mask_dt_out parameter and
27573         use it to pass back the definition type.
27574         (vect_check_store_rhs): Likewise rhs_dt_out.
27575         (vect_build_gather_load_calls): Add a mask_dt argument and use
27576         it instead of a call to vect_is_simple_use.
27577         (vectorizable_store): Update calls to vect_check_load_store_mask
27578         and vect_check_store_rhs.  Use the dt returned by the latter instead
27579         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
27580         instead of calls to vect_is_simple_use.  Pass the scalar rather
27581         than the vector operand to vect_is_simple_use when handling
27582         second and subsequent copies of an rhs value.
27583         (vectorizable_load): Update calls to vect_check_load_store_mask
27584         and vect_build_gather_load_calls.  Use the cached mask_dt and
27585         gs_info.offset_dt instead of calls to vect_is_simple_use.
27587 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
27589         PR middle-end/83945
27590         * tree-emutls.c: Include gimplify.h.
27591         (lower_emutls_2): New function.
27592         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
27593         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
27594         it before further processing.
27596         PR target/83930
27597         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
27598         UINTVAL (trueop1) instead of INTVAL (op1).
27600 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
27602         PR debug/81570
27603         PR debug/83728
27604         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
27605         INCOMING_FRAME_SP_OFFSET if not defined.
27606         (scan_trace): Add ENTRY argument.  If true and
27607         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
27608         emit a note to adjust the CFA offset.
27609         (create_cfi_notes): Adjust scan_trace callers.
27610         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
27611         INCOMING_FRAME_SP_OFFSET in the CIE.
27612         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
27613         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
27614         Likewise.
27615         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
27616         * doc/tm.texi: Regenerated.
27618 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
27620         PR rtl-optimization/83147
27621         * lra-constraints.c (remove_inheritance_pseudos): Use
27622         lra_substitute_pseudo_within_insn.
27624 2018-01-19  Tom de Vries  <tom@codesourcery.com>
27625             Cesar Philippidis  <cesar@codesourcery.com>
27627         PR target/83920
27628         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
27630 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
27632         PR target/83790
27633         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
27634         spaces for function labels.
27636 2018-01-19  Martin Liska  <mliska@suse.cz>
27638         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
27639         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
27640         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
27641         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
27642         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
27643         (PRED_CONST_RETURN): Change from 69 to 65.
27644         (PRED_NULL_RETURN): Change from 91 to 71.
27645         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
27646         (PRED_LOOP_GUARD): Change from 66 to 73.
27648 2018-01-19  Martin Liska  <mliska@suse.cz>
27650         * predict.c (predict_insn_def): Add new assert.
27651         (struct branch_predictor): Change type to signed integer.
27652         (test_prediction_value_range): Amend test to cover
27653         PROB_UNINITIALIZED.
27654         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
27655         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
27656         (PRED_LOOP_ITERATIONS_MAX): Likewise.
27657         (PRED_LOOP_IV_COMPARE): Likewise.
27658         * predict.h (PROB_UNINITIALIZED): Define new constant.
27660 2018-01-19  Martin Liska  <mliska@suse.cz>
27662         * predict.c (dump_prediction): Add new format for
27663         analyze_brprob.py script which is enabled with -details
27664         suboption.
27665         * profile-count.h (precise_p): New function.
27667 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
27669         PR tree-optimization/83922
27670         * tree-vect-loop.c (vect_verify_full_masking): Return false if
27671         there are no statements that need masking.
27672         (vect_active_double_reduction_p): New function.
27673         (vect_analyze_loop_operations): Use it when handling phis that
27674         are not in the loop header.
27676 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
27678         PR tree-optimization/83914
27679         * tree-vect-loop.c (vectorizable_induction): Don't convert
27680         init_expr or apply the peeling adjustment for inductions
27681         that are nested within the vectorized loop.
27683 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27685         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
27686         instead of NEG.
27688 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
27690         PR sanitizer/81715
27691         PR testsuite/83882
27692         * function.h (gimplify_parameters): Add gimple_seq * argument.
27693         * function.c: Include gimple.h and options.h.
27694         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
27695         for the added local temporaries if needed.
27696         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
27697         if there are any parameter cleanups, wrap whole body into a
27698         try/finally with the cleanups.
27700 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
27702         PR target/82964
27703         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27704         Use GET_MODE_CLASS for scalar floating point.
27706 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
27708         PR ipa/82256
27709         patch by PaX Team
27710         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
27711         Fix call of call_cgraph_insertion_hooks.
27713 2018-01-18  Martin Sebor  <msebor@redhat.com>
27715         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
27717 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
27719         PR ipa/83619
27720         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
27721         frequencies.
27723 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
27725         PR other/70268
27726         * common.opt: (-ffile-prefix-map): New option.
27727         * opts.c (common_handle_option): Defer it.
27728         * opts-global.c (handle_common_deferred_options): Handle it.
27729         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
27730         * file-prefix-map.h: New file.
27731         (remap_debug_filename, add_debug_prefix_map): ...here.
27732         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
27733         * final.c (debug_prefix_map, add_debug_prefix_map
27734         remap_debug_filename): Move to...
27735         * file-prefix-map.c: New file.
27736         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
27737         generalize, get rid of alloca(), use strrchr() instead of strchr().
27738         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
27739         Implement in terms of add_prefix_map().
27740         (remap_macro_filename, remap_debug_filename): Implement in term of
27741         remap_filename().
27742         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
27743         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
27744         * dbxout.c: Include file-prefix-map.h.
27745         * varasm.c: Likewise.
27746         * vmsdbgout.c: Likewise.
27747         * xcoffout.c: Likewise.
27748         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
27749         * doc/cppopts.texi (-fmacro-prefix-map): Document.
27750         * doc/invoke.texi (-ffile-prefix-map): Document.
27751         (-fdebug-prefix-map): Update description.
27753 2018-01-18  Martin Liska  <mliska@suse.cz>
27755         * config/i386/i386.c (indirect_thunk_name): Document that also
27756         lfence is emitted.
27757         (output_indirect_thunk): Document why both instructions
27758         (pause and lfence) are generated.
27760 2018-01-18  Richard Biener  <rguenther@suse.de>
27762         PR tree-optimization/83887
27763         * graphite-scop-detection.c
27764         (scop_detection::get_nearest_dom_with_single_entry): Remove.
27765         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
27766         (scop_detection::merge_sese): Re-implement with a flood-fill
27767         algorithm that properly finds a SESE region if it exists.
27769 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
27771         PR c/61240
27772         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
27773         pointer_diff optimizations use view_convert instead of convert.
27775 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27777         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27778         Generate different code for -mno-speculate-indirect-jumps.
27779         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27780         (*call_indirect_aix<mode>): Disable for
27781         -mno-speculate-indirect-jumps.
27782         (*call_indirect_aix<mode>_nospec): New define_insn.
27783         (*call_value_indirect_aix<mode>): Disable for
27784         -mno-speculate-indirect-jumps.
27785         (*call_value_indirect_aix<mode>_nospec): New define_insn.
27786         (*sibcall_nonlocal_sysv<mode>): Generate different code for
27787         -mno-speculate-indirect-jumps.
27788         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27790 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
27792         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
27793         long double type, set the flags for noting the default long double
27794         type, even if we don't pass or return a long double type.
27796 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
27798         PR ipa/83051
27799         * ipa-inline.c (flatten_function): Do not overwrite final inlining
27800         failure.
27802 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
27804         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
27805         support for merge[hl].
27806         (fold_mergehl_helper): New helper function.
27807         (tree-vector-builder.h): New #include for tree_vector_builder usage.
27808         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
27809         (altivec_vmrglw_direct): Add xxmrglw insn.
27811 2018-01-17  Andrew Waterman  <andrew@sifive.com>
27813         * config/riscv/riscv.c (riscv_conditional_register_usage): If
27814         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
27816 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
27818         PR lto/83121
27819         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
27820         call the lto_location_cache before reading the
27821         DECL_SOURCE_LOCATION of the types.
27823 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
27824             Richard Sandiford  <richard.sandiford@linaro.org>
27826         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
27827         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
27828         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
27829         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
27830         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
27831         Add declaration.
27832         * config/aarch64/constraints.md (aarch64_movti_operand):
27833         Limit immediates.
27834         * config/aarch64/predicates.md (Uti): Add new constraint.
27836 2018-01-17  Carl Love  <cel@us.ibm.com>
27838         * config/rs6000/vsx.md (define_expand xl_len_r,
27839         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
27840         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
27841         lxvll.
27842         (define_expand, define_insn): Move the shift left from  the
27843         define_insn to the define_expand for lxvl and stxvl instructions.
27844         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
27845         and XL_LEN_R definitions to PURE.
27847 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
27849         * config/i386/i386.c (indirect_thunk_name): Declare regno
27850         as unsigned int.  Compare regno with INVALID_REGNUM.
27851         (output_indirect_thunk): Ditto.
27852         (output_indirect_thunk_function): Ditto.
27853         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
27854         in the call to output_indirect_thunk_function.
27856 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
27858         PR middle-end/83884
27859         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
27860         rather than the size of inner_type to determine the stack slot size
27861         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
27863 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
27865         PR target/83546
27866         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
27867         to PTA_SILVERMONT.
27869 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
27871         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
27872         endian Linux systems to optionally enable multilibs for selecting
27873         the long double type if the user configured an explicit type.
27874         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
27875         have no long double multilibs if not defined.
27876         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
27877         warn if the user used -mabi={ieee,ibm}longdouble and we built
27878         multilibs for long double.
27879         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
27880         appropriate multilib option.
27881         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
27882         multilib options.
27883         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
27884         for building long double multilibs.
27885         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
27887 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
27889         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
27890         copies.
27892         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
27893         64 bits.
27894         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
27895         128 bits.
27897         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
27898         variables.
27900         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
27901         return value.
27903 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
27905         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
27906         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
27908 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
27910         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
27911         different rtl trees depending on TARGET_64BIT.
27912         (rs6000_gen_lvx): Likewise.
27914 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
27916         * config/visium/visium.md (nop): Tweak comment.
27917         (hazard_nop): Likewise.
27919 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27921         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
27922         -mspeculate-indirect-jumps.
27923         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
27924         for -mno-speculate-indirect-jumps.
27925         (*call_indirect_elfv2<mode>_nospec): New define_insn.
27926         (*call_value_indirect_elfv2<mode>): Disable for
27927         -mno-speculate-indirect-jumps.
27928         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
27929         (indirect_jump): Emit different RTL for
27930         -mno-speculate-indirect-jumps.
27931         (*indirect_jump<mode>): Disable for
27932         -mno-speculate-indirect-jumps.
27933         (*indirect_jump<mode>_nospec): New define_insn.
27934         (tablejump): Emit different RTL for
27935         -mno-speculate-indirect-jumps.
27936         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
27937         (tablejumpsi_nospec): New define_expand.
27938         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
27939         (tablejumpdi_nospec): New define_expand.
27940         (*tablejump<mode>_internal1): Disable for
27941         -mno-speculate-indirect-jumps.
27942         (*tablejump<mode>_internal1_nospec): New define_insn.
27943         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
27944         option.
27946 2018-01-16  Artyom Skrobov tyomitch@gmail.com
27948         * caller-save.c (insert_save): Drop unnecessary parameter.  All
27949         callers updated.
27951 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
27952             Richard Biener  <rguenth@suse.de>
27954         PR libgomp/83590
27955         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
27956         return early, inline manually is_gimple_sizepos.  Make sure if we
27957         call gimplify_expr we don't end up with a gimple constant.
27958         * tree.c (variably_modified_type_p): Don't return true for
27959         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
27960         * gimplify.h (is_gimple_sizepos): Remove.
27962 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
27964         PR tree-optimization/83857
27965         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
27966         vectorizable_live_operation for pure SLP statements.
27967         (vectorizable_live_operation): Handle PHIs.
27969 2018-01-16  Richard Biener  <rguenther@suse.de>
27971         PR tree-optimization/83867
27972         * tree-vect-stmts.c (vect_transform_stmt): Precompute
27973         nested_in_vect_loop_p since the scalar stmt may get invalidated.
27975 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
27977         PR c/83844
27978         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
27979         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
27980         If off is not INTEGER_CST, issue a may not be aligned warning
27981         rather than isn't aligned.  Use isn%'t rather than isn't.
27982         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
27983         into MULT_EXPR.
27984         <case MULT_EXPR>: Improve the case when bottom and one of the
27985         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
27986         operand, in that case check if the other operand is multiple of
27987         bottom divided by the INTEGER_CST operand.
27989 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
27991         PR target/83858
27992         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
27993         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
27994         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
27995         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
27996         * config/pa/pa.c (pa_function_arg_advance): Likewise.
27997         (pa_function_arg, pa_arg_partial_bytes): Likewise.
27998         (pa_function_arg_size): New function.
28000 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
28002         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
28003         in a separate statement.
28005 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
28007         PR tree-optimization/83847
28008         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
28009         group gathers and scatters.
28011 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
28013         PR rtl-optimization/86620
28014         * params.def (max-sched-ready-insns): Bump minimum value to 1.
28016         PR rtl-optimization/83213
28017         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
28018         to last if both are JUMP_INSNs.
28020         PR tree-optimization/83843
28021         * gimple-ssa-store-merging.c
28022         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
28023         store_immediate_info for bswap/nop orig_stores.
28025 2018-01-15  Andrew Waterman  <andrew@sifive.com>
28027         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
28028         !TARGET_MUL.
28029         <UDIV>: Increase cost if !TARGET_DIV.
28031 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
28033         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
28034         (define_attr "cr_logical_3op"): New.
28035         (cceq_ior_compare): Adjust.
28036         (cceq_ior_compare_complement): Adjust.
28037         (*cceq_rev_compare): Adjust.
28038         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
28039         (is_cracked_insn): Adjust.
28040         (insn_must_be_first_in_group): Adjust.
28041         * config/rs6000/40x.md: Adjust.
28042         * config/rs6000/440.md: Adjust.
28043         * config/rs6000/476.md: Adjust.
28044         * config/rs6000/601.md: Adjust.
28045         * config/rs6000/603.md: Adjust.
28046         * config/rs6000/6xx.md: Adjust.
28047         * config/rs6000/7450.md: Adjust.
28048         * config/rs6000/7xx.md: Adjust.
28049         * config/rs6000/8540.md: Adjust.
28050         * config/rs6000/cell.md: Adjust.
28051         * config/rs6000/e300c2c3.md: Adjust.
28052         * config/rs6000/e500mc.md: Adjust.
28053         * config/rs6000/e500mc64.md: Adjust.
28054         * config/rs6000/e5500.md: Adjust.
28055         * config/rs6000/e6500.md: Adjust.
28056         * config/rs6000/mpc.md: Adjust.
28057         * config/rs6000/power4.md: Adjust.
28058         * config/rs6000/power5.md: Adjust.
28059         * config/rs6000/power6.md: Adjust.
28060         * config/rs6000/power7.md: Adjust.
28061         * config/rs6000/power8.md: Adjust.
28062         * config/rs6000/power9.md: Adjust.
28063         * config/rs6000/rs64.md: Adjust.
28064         * config/rs6000/titan.md: Adjust.
28066 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28068         * config/i386/predicates.md (indirect_branch_operand): Rewrite
28069         ix86_indirect_branch_register logic.
28071 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28073         * config/i386/constraints.md (Bs): Update
28074         ix86_indirect_branch_register check.  Don't check
28075         ix86_indirect_branch_register with GOT_memory_operand.
28076         (Bw): Likewise.
28077         * config/i386/predicates.md (GOT_memory_operand): Don't check
28078         ix86_indirect_branch_register here.
28079         (GOT32_symbol_operand): Likewise.
28081 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28083         * config/i386/predicates.md (constant_call_address_operand):
28084         Rewrite ix86_indirect_branch_register logic.
28085         (sibcall_insn_operand): Likewise.
28087 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28089         * config/i386/constraints.md (Bs): Replace
28090         ix86_indirect_branch_thunk_register with
28091         ix86_indirect_branch_register.
28092         (Bw): Likewise.
28093         * config/i386/i386.md (indirect_jump): Likewise.
28094         (tablejump): Likewise.
28095         (*sibcall_memory): Likewise.
28096         (*sibcall_value_memory): Likewise.
28097         Peepholes of indirect call and jump via memory: Likewise.
28098         * config/i386/i386.opt: Likewise.
28099         * config/i386/predicates.md (indirect_branch_operand): Likewise.
28100         (GOT_memory_operand): Likewise.
28101         (call_insn_operand): Likewise.
28102         (sibcall_insn_operand): Likewise.
28103         (GOT32_symbol_operand): Likewise.
28105 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
28107         PR middle-end/83837
28108         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
28109         type rather than type addr's type points to.
28110         (expand_omp_atomic_mutex): Likewise.
28111         (expand_omp_atomic): Likewise.
28113 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28115         PR target/83839
28116         * config/i386/i386.c (output_indirect_thunk_function): Use
28117         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
28118         for  __x86_return_thunk.
28120 2018-01-15  Richard Biener  <rguenther@suse.de>
28122         PR middle-end/83850
28123         * expmed.c (extract_bit_field_1): Fix typo.
28125 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28127         PR target/83687
28128         * config/arm/iterators.md (VF): New mode iterator.
28129         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
28130         Remove integer-related logic from pattern.
28131         (neon_vabd<mode>_3): Likewise.
28133 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
28135         PR middle-end/82694
28136         * common.opt (fstrict-overflow): No longer an alias.
28137         (fwrapv-pointer): New option.
28138         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
28139         also for pointer types based on flag_wrapv_pointer.
28140         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
28141         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
28142         opts->x_flag_wrapv got set.
28143         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
28144         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
28145         POINTER_TYPE_OVERFLOW_UNDEFINED.
28146         * match.pd: Likewise in address comparison pattern.
28147         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
28149 2018-01-15  Richard Biener  <rguenther@suse.de>
28151         PR lto/83804
28152         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
28153         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
28154         Reset type names to their identifier if their TYPE_DECL doesn't
28155         have linkage (and thus is used for ODR and devirt).
28156         (save_debug_info_for_decl): Remove.
28157         (save_debug_info_for_type): Likewise.
28158         (add_tree_to_fld_list): Adjust.
28159         * tree-pretty-print.c (dump_generic_node): Make dumping of
28160         type names more robust.
28162 2018-01-15  Richard Biener  <rguenther@suse.de>
28164         * BASE-VER: Bump to 8.0.1.
28166 2018-01-14  Martin Sebor  <msebor@redhat.com>
28168         PR other/83508
28169         * builtins.c (check_access): Avoid warning when the no-warning bit
28170         is set.
28172 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
28174         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
28175         * ira-color (allocno_hard_regs_compare): Likewise.
28177 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
28179         PR target/83013
28180         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
28181         Use .pushsection/.popsection.
28183 2018-01-14  Martin Sebor  <msebor@redhat.com>
28185         PR c++/81327
28186         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
28188 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
28190         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
28191         entry from extra_headers.
28192         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
28193         extra_headers, make the list bitwise identical to the i?86-*-* one.
28195 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28197         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
28198         -mcmodel=large with -mindirect-branch=thunk,
28199         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
28200         -mfunction-return=thunk-extern.
28201         * doc/invoke.texi: Document -mcmodel=large is incompatible with
28202         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
28203         -mfunction-return=thunk and -mfunction-return=thunk-extern.
28205 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28207         * config/i386/i386.c (print_reg): Print the name of the full
28208         integer register without '%'.
28209         (ix86_print_operand): Handle 'V'.
28210         * doc/extend.texi: Document 'V' modifier.
28212 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28214         * config/i386/constraints.md (Bs): Disallow memory operand for
28215         -mindirect-branch-register.
28216         (Bw): Likewise.
28217         * config/i386/predicates.md (indirect_branch_operand): Likewise.
28218         (GOT_memory_operand): Likewise.
28219         (call_insn_operand): Likewise.
28220         (sibcall_insn_operand): Likewise.
28221         (GOT32_symbol_operand): Likewise.
28222         * config/i386/i386.md (indirect_jump): Call convert_memory_address
28223         for -mindirect-branch-register.
28224         (tablejump): Likewise.
28225         (*sibcall_memory): Likewise.
28226         (*sibcall_value_memory): Likewise.
28227         Disallow peepholes of indirect call and jump via memory for
28228         -mindirect-branch-register.
28229         (*call_pop): Replace m with Bw.
28230         (*call_value_pop): Likewise.
28231         (*sibcall_pop_memory): Replace m with Bs.
28232         * config/i386/i386.opt (mindirect-branch-register): New option.
28233         * doc/invoke.texi: Document -mindirect-branch-register option.
28235 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28237         * config/i386/i386-protos.h (ix86_output_function_return): New.
28238         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
28239         set function_return_type.
28240         (indirect_thunk_name): Add ret_p to indicate thunk for function
28241         return.
28242         (output_indirect_thunk_function): Pass false to
28243         indirect_thunk_name.
28244         (ix86_output_indirect_branch_via_reg): Likewise.
28245         (ix86_output_indirect_branch_via_push): Likewise.
28246         (output_indirect_thunk_function): Create alias for function
28247         return thunk if regno < 0.
28248         (ix86_output_function_return): New function.
28249         (ix86_handle_fndecl_attribute): Handle function_return.
28250         (ix86_attribute_table): Add function_return.
28251         * config/i386/i386.h (machine_function): Add
28252         function_return_type.
28253         * config/i386/i386.md (simple_return_internal): Use
28254         ix86_output_function_return.
28255         (simple_return_internal_long): Likewise.
28256         * config/i386/i386.opt (mfunction-return=): New option.
28257         (indirect_branch): Mention -mfunction-return=.
28258         * doc/extend.texi: Document function_return function attribute.
28259         * doc/invoke.texi: Document -mfunction-return= option.
28261 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28263         * config/i386/i386-opts.h (indirect_branch): New.
28264         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
28265         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
28266         with local indirect jump when converting indirect call and jump.
28267         (ix86_set_indirect_branch_type): New.
28268         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
28269         (indirectlabelno): New.
28270         (indirect_thunk_needed): Likewise.
28271         (indirect_thunk_bnd_needed): Likewise.
28272         (indirect_thunks_used): Likewise.
28273         (indirect_thunks_bnd_used): Likewise.
28274         (INDIRECT_LABEL): Likewise.
28275         (indirect_thunk_name): Likewise.
28276         (output_indirect_thunk): Likewise.
28277         (output_indirect_thunk_function): Likewise.
28278         (ix86_output_indirect_branch_via_reg): Likewise.
28279         (ix86_output_indirect_branch_via_push): Likewise.
28280         (ix86_output_indirect_branch): Likewise.
28281         (ix86_output_indirect_jmp): Likewise.
28282         (ix86_code_end): Call output_indirect_thunk_function if needed.
28283         (ix86_output_call_insn): Call ix86_output_indirect_branch if
28284         needed.
28285         (ix86_handle_fndecl_attribute): Handle indirect_branch.
28286         (ix86_attribute_table): Add indirect_branch.
28287         * config/i386/i386.h (machine_function): Add indirect_branch_type
28288         and has_local_indirect_jump.
28289         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
28290         to true.
28291         (tablejump): Likewise.
28292         (*indirect_jump): Use ix86_output_indirect_jmp.
28293         (*tablejump_1): Likewise.
28294         (simple_return_indirect_internal): Likewise.
28295         * config/i386/i386.opt (mindirect-branch=): New option.
28296         (indirect_branch): New.
28297         (keep): Likewise.
28298         (thunk): Likewise.
28299         (thunk-inline): Likewise.
28300         (thunk-extern): Likewise.
28301         * doc/extend.texi: Document indirect_branch function attribute.
28302         * doc/invoke.texi: Document -mindirect-branch= option.
28304 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
28306         PR ipa/83051
28307         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
28309 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
28311         * ipa-inline.c (want_inline_small_function_p): Return false if
28312         inlining has already failed with CIF_FINAL_ERROR.
28313         (update_caller_keys): Call want_inline_small_function_p before
28314         can_inline_edge_p.
28315         (update_callee_keys): Likewise.
28317 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
28319         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
28320         New function.
28321         (rs6000_quadword_masked_address_p): Likewise.
28322         (quad_aligned_load_p): Likewise.
28323         (quad_aligned_store_p): Likewise.
28324         (const_load_sequence_p): Add comment to describe the outer-most loop.
28325         (mimic_memory_attributes_and_flags): New function.
28326         (rs6000_gen_stvx): Likewise.
28327         (replace_swapped_aligned_store): Likewise.
28328         (rs6000_gen_lvx): Likewise.
28329         (replace_swapped_aligned_load): Likewise.
28330         (replace_swapped_load_constant): Capitalize argument name in
28331         comment describing this function.
28332         (rs6000_analyze_swaps): Add a third pass to search for vector loads
28333         and stores that access quad-word aligned addresses and replace
28334         with stvx or lvx instructions when appropriate.
28335         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
28336         New function prototype.
28337         (rs6000_quadword_masked_address_p): Likewise.
28338         (rs6000_gen_lvx): Likewise.
28339         (rs6000_gen_stvx): Likewise.
28340         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
28341         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
28342         when memory address is aligned.
28343         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
28344         this split to select lvx instruction when memory address is aligned.
28345         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
28346         instruction when memory address is aligned.
28347         (*vsx_le_perm_load_v16qi): Likewise.
28348         (four unnamed splitters): Modify to select the stvx instruction
28349         when memory is aligned.
28351 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
28353         * predict.c (determine_unlikely_bbs): Handle correctly BBs
28354         which appears in the queue multiple times.
28356 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28357             Alan Hayward  <alan.hayward@arm.com>
28358             David Sherwood  <david.sherwood@arm.com>
28360         * tree-vectorizer.h (vec_lower_bound): New structure.
28361         (_loop_vec_info): Add check_nonzero and lower_bounds.
28362         (LOOP_VINFO_CHECK_NONZERO): New macro.
28363         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
28364         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
28365         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
28366         fields.  Make seg_len the distance travelled, not including the
28367         access size.
28368         (dr_direction_indicator): Declare.
28369         (dr_zero_step_indicator): Likewise.
28370         (dr_known_forward_stride_p): Likewise.
28371         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
28372         tree-ssanames.h.
28373         (runtime_alias_check_p): Allow runtime alias checks with
28374         variable strides.
28375         (operator ==): Compare access_size and align.
28376         (prune_runtime_alias_test_list): Rework for new distinction between
28377         the access_size and seg_len.
28378         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
28379         segment lengths.
28380         (get_segment_min_max): New function.
28381         (create_intersect_range_checks): Use it.
28382         (dr_step_indicator): New function.
28383         (dr_direction_indicator): Likewise.
28384         (dr_zero_step_indicator): Likewise.
28385         (dr_known_forward_stride_p): Likewise.
28386         * tree-loop-distribution.c (data_ref_segment_size): Return
28387         DR_STEP * (niters - 1).
28388         (compute_alias_check_pairs): Update call to the dr_with_seg_len
28389         constructor.
28390         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
28391         (vect_preserves_scalar_order_p): New function, split out from...
28392         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
28393         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
28394         (vect_vfa_access_size): New function.
28395         (vect_vfa_align): Likewise.
28396         (vect_compile_time_alias): Take access_size_a and access_b arguments.
28397         (dump_lower_bound): New function.
28398         (vect_check_lower_bound): Likewise.
28399         (vect_small_gap_p): Likewise.
28400         (vectorizable_with_step_bound_p): Likewise.
28401         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
28402         depencies if the vectorization factor is 1.  Convert the checks
28403         for nonzero steps into checks on the bounds of DR_STEP.  Try using
28404         a bunds check for variable steps if the minimum required step is
28405         relatively small. Update calls to the dr_with_seg_len
28406         constructor and to vect_compile_time_alias.
28407         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
28408         function.
28409         (vect_loop_versioning): Call it.
28410         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
28411         when retrying.
28412         (vect_estimate_min_profitable_iters): Account for any bounds checks.
28414 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28415             Alan Hayward  <alan.hayward@arm.com>
28416             David Sherwood  <david.sherwood@arm.com>
28418         * doc/sourcebuild.texi (vect_scatter_store): Document.
28419         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
28420         optabs.
28421         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
28422         Document.
28423         * genopinit.c (main): Add supports_vec_scatter_store and
28424         supports_vec_scatter_store_cached to target_optabs.
28425         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
28426         IFN_MASK_SCATTER_STORE.
28427         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
28428         functions.
28429         * internal-fn.h (internal_store_fn_p): Declare.
28430         (internal_fn_stored_value_index): Likewise.
28431         * internal-fn.c (scatter_store_direct): New macro.
28432         (expand_scatter_store_optab_fn): New function.
28433         (direct_scatter_store_optab_supported_p): New macro.
28434         (internal_store_fn_p): New function.
28435         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
28436         IFN_MASK_SCATTER_STORE.
28437         (internal_fn_mask_index): Likewise.
28438         (internal_fn_stored_value_index): New function.
28439         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
28440         for scatter stores.
28441         * optabs-query.h (supports_vec_scatter_store_p): Declare.
28442         * optabs-query.c (supports_vec_scatter_store_p): New function.
28443         * tree-vectorizer.h (vect_get_store_rhs): Declare.
28444         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
28445         true for scatter stores.
28446         (vect_gather_scatter_fn_p): Handle scatter stores too.
28447         (vect_check_gather_scatter): Consider using scatter stores if
28448         supports_vec_scatter_store_p.
28449         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
28450         scatter stores too.
28451         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
28452         internal_fn_stored_value_index.
28453         (check_load_store_masking): Handle scatter stores too.
28454         (vect_get_store_rhs): Make public.
28455         (vectorizable_call): Use internal_store_fn_p.
28456         (vectorizable_store): Handle scatter store internal functions.
28457         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
28458         when deciding whether the end of the group has been reached.
28459         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
28460         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
28461         (mask_scatter_store<mode>): New insns.
28463 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28464             Alan Hayward  <alan.hayward@arm.com>
28465             David Sherwood  <david.sherwood@arm.com>
28467         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
28468         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
28469         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
28470         function.
28471         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
28472         Use vect_truncate_gather_scatter_offset if we can't treat the
28473         operation as a normal gather load or scatter store.
28474         (get_group_load_store_type): Take the gather_scatter_info
28475         as argument.  Try using a gather load or scatter store for
28476         single-element groups.
28477         (get_load_store_type): Update calls to get_group_load_store_type
28478         and vect_use_strided_gather_scatters_p.
28480 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28481             Alan Hayward  <alan.hayward@arm.com>
28482             David Sherwood  <david.sherwood@arm.com>
28484         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
28485         optional tree argument.
28486         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
28487         null target hooks.
28488         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
28489         but continue to use the current value as a fallback.
28490         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
28491         to compare the updates.
28492         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
28493         (get_load_store_type): Use it when handling a strided access.
28494         (vect_get_strided_load_store_ops): New function.
28495         (vect_get_data_ptr_increment): Likewise.
28496         (vectorizable_load): Handle strided gather loads.  Always pass
28497         a step to vect_create_data_ref_ptr and bump_vector_ptr.
28499 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28500             Alan Hayward  <alan.hayward@arm.com>
28501             David Sherwood  <david.sherwood@arm.com>
28503         * doc/md.texi (gather_load@var{m}): Document.
28504         (mask_gather_load@var{m}): Likewise.
28505         * genopinit.c (main): Add supports_vec_gather_load and
28506         supports_vec_gather_load_cached to target_optabs.
28507         * optabs-tree.c (init_tree_optimization_optabs): Use
28508         ggc_cleared_alloc to allocate target_optabs.
28509         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
28510         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
28511         functions.
28512         * internal-fn.h (internal_load_fn_p): Declare.
28513         (internal_gather_scatter_fn_p): Likewise.
28514         (internal_fn_mask_index): Likewise.
28515         (internal_gather_scatter_fn_supported_p): Likewise.
28516         * internal-fn.c (gather_load_direct): New macro.
28517         (expand_gather_load_optab_fn): New function.
28518         (direct_gather_load_optab_supported_p): New macro.
28519         (direct_internal_fn_optab): New function.
28520         (internal_load_fn_p): Likewise.
28521         (internal_gather_scatter_fn_p): Likewise.
28522         (internal_fn_mask_index): Likewise.
28523         (internal_gather_scatter_fn_supported_p): Likewise.
28524         * optabs-query.c (supports_at_least_one_mode_p): New function.
28525         (supports_vec_gather_load_p): Likewise.
28526         * optabs-query.h (supports_vec_gather_load_p): Declare.
28527         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
28528         and memory_type field.
28529         (NUM_PATTERNS): Bump to 15.
28530         * tree-vect-data-refs.c: Include internal-fn.h.
28531         (vect_gather_scatter_fn_p): New function.
28532         (vect_describe_gather_scatter_call): Likewise.
28533         (vect_check_gather_scatter): Try using internal functions for
28534         gather loads.  Recognize existing calls to a gather load function.
28535         (vect_analyze_data_refs): Consider using gather loads if
28536         supports_vec_gather_load_p.
28537         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
28538         (vect_get_gather_scatter_offset_type): Likewise.
28539         (vect_convert_mask_for_vectype): Likewise.
28540         (vect_add_conversion_to_patterm): Likewise.
28541         (vect_try_gather_scatter_pattern): Likewise.
28542         (vect_recog_gather_scatter_pattern): New pattern recognizer.
28543         (vect_vect_recog_func_ptrs): Add it.
28544         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
28545         internal_fn_mask_index and internal_gather_scatter_fn_p.
28546         (check_load_store_masking): Take the gather_scatter_info as an
28547         argument and handle gather loads.
28548         (vect_get_gather_scatter_ops): New function.
28549         (vectorizable_call): Check internal_load_fn_p.
28550         (vectorizable_load): Likewise.  Handle gather load internal
28551         functions.
28552         (vectorizable_store): Update call to check_load_store_masking.
28553         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
28554         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
28555         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
28556         (aarch64_gather_scale_operand_d): New predicates.
28557         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
28558         (mask_gather_load<mode>): New insns.
28560 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28561             Alan Hayward  <alan.hayward@arm.com>
28562             David Sherwood  <david.sherwood@arm.com>
28564         * optabs.def (fold_left_plus_optab): New optab.
28565         * doc/md.texi (fold_left_plus_@var{m}): Document.
28566         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
28567         * internal-fn.c (fold_left_direct): Define.
28568         (expand_fold_left_optab_fn): Likewise.
28569         (direct_fold_left_optab_supported_p): Likewise.
28570         * fold-const-call.c (fold_const_fold_left): New function.
28571         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
28572         * tree-parloops.c (valid_reduction_p): New function.
28573         (gather_scalar_reductions): Use it.
28574         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
28575         (vect_finish_replace_stmt): Declare.
28576         * tree-vect-loop.c (fold_left_reduction_fn): New function.
28577         (needs_fold_left_reduction_p): New function, split out from...
28578         (vect_is_simple_reduction): ...here.  Accept reductions that
28579         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
28580         (vect_force_simple_reduction): Also store the reduction type in
28581         the assignment's STMT_VINFO_REDUC_TYPE.
28582         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
28583         (merge_with_identity): New function.
28584         (vect_expand_fold_left): Likewise.
28585         (vectorize_fold_left_reduction): Likewise.
28586         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
28587         scalar phi in place for it.  Check for target support and reject
28588         cases that would reassociate the operation.  Defer the transform
28589         phase to vectorize_fold_left_reduction.
28590         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
28591         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
28592         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
28594 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28596         * tree-if-conv.c (predicate_mem_writes): Remove redundant
28597         call to ifc_temp_var.
28599 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28600             Alan Hayward  <alan.hayward@arm.com>
28601             David Sherwood  <david.sherwood@arm.com>
28603         * target.def (legitimize_address_displacement): Take the original
28604         offset as a poly_int.
28605         * targhooks.h (default_legitimize_address_displacement): Update
28606         accordingly.
28607         * targhooks.c (default_legitimize_address_displacement): Likewise.
28608         * doc/tm.texi: Regenerate.
28609         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
28610         as an argument, moving assert of ad->disp == ad->disp_term to...
28611         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
28612         Try calling targetm.legitimize_address_displacement before expanding
28613         the address rather than afterwards, and adjust for the new interface.
28614         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
28615         Match the new hook interface.  Handle SVE addresses.
28616         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
28617         new hook interface.
28619 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28621         * Makefile.in (OBJS): Add early-remat.o.
28622         * target.def (select_early_remat_modes): New hook.
28623         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
28624         * doc/tm.texi: Regenerate.
28625         * targhooks.h (default_select_early_remat_modes): Declare.
28626         * targhooks.c (default_select_early_remat_modes): New function.
28627         * timevar.def (TV_EARLY_REMAT): New timevar.
28628         * passes.def (pass_early_remat): New pass.
28629         * tree-pass.h (make_pass_early_remat): Declare.
28630         * early-remat.c: New file.
28631         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
28632         function.
28633         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
28635 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28636             Alan Hayward  <alan.hayward@arm.com>
28637             David Sherwood  <david.sherwood@arm.com>
28639         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
28640         vfm1 with a bound_epilog parameter.
28641         (vect_do_peeling): Update calls accordingly, and move the prologue
28642         call earlier in the function.  Treat the base bound_epilog as 0 for
28643         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
28644         this base when peeling for gaps.
28645         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
28646         with fully-masked loops.
28647         (vect_estimate_min_profitable_iters): Handle the single peeled
28648         iteration in that case.
28650 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28651             Alan Hayward  <alan.hayward@arm.com>
28652             David Sherwood  <david.sherwood@arm.com>
28654         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
28655         single-element interleaving even if the size is not a power of 2.
28656         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
28657         accesses for single-element interleaving if the group size is
28658         not a power of 2.
28660 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28661             Alan Hayward  <alan.hayward@arm.com>
28662             David Sherwood  <david.sherwood@arm.com>
28664         * doc/md.texi (fold_extract_last_@var{m}): Document.
28665         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
28666         * optabs.def (fold_extract_last_optab): New optab.
28667         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
28668         * internal-fn.c (fold_extract_direct): New macro.
28669         (expand_fold_extract_optab_fn): Likewise.
28670         (direct_fold_extract_optab_supported_p): Likewise.
28671         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
28672         * tree-vect-loop.c (vect_model_reduction_cost): Handle
28673         EXTRACT_LAST_REDUCTION.
28674         (get_initial_def_for_reduction): Do not create an initial vector
28675         for EXTRACT_LAST_REDUCTION reductions.
28676         (vectorizable_reduction): Leave the scalar phi in place for
28677         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
28678         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
28679         epilogue code for EXTRACT_LAST_REDUCTION and defer the
28680         transform phase to vectorizable_condition.
28681         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
28682         split out from...
28683         (vect_finish_stmt_generation): ...here.
28684         (vect_finish_replace_stmt): New function.
28685         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
28686         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
28687         pattern.
28688         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
28690 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28691             Alan Hayward  <alan.hayward@arm.com>
28692             David Sherwood  <david.sherwood@arm.com>
28694         * doc/md.texi (extract_last_@var{m}): Document.
28695         * optabs.def (extract_last_optab): New optab.
28696         * internal-fn.def (EXTRACT_LAST): New internal function.
28697         * internal-fn.c (cond_unary_direct): New macro.
28698         (expand_cond_unary_optab_fn): Likewise.
28699         (direct_cond_unary_optab_supported_p): Likewise.
28700         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
28701         loops using EXTRACT_LAST.
28702         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
28703         (extract_last_<mode>): ...this optab.
28704         (vec_extract<mode><Vel>): Update accordingly.
28706 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28707             Alan Hayward  <alan.hayward@arm.com>
28708             David Sherwood  <david.sherwood@arm.com>
28710         * target.def (empty_mask_is_expensive): New hook.
28711         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
28712         * doc/tm.texi: Regenerate.
28713         * targhooks.h (default_empty_mask_is_expensive): Declare.
28714         * targhooks.c (default_empty_mask_is_expensive): New function.
28715         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
28716         if the target says that empty masks are expensive.
28717         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
28718         New function.
28719         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
28721 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28722             Alan Hayward  <alan.hayward@arm.com>
28723             David Sherwood  <david.sherwood@arm.com>
28725         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
28726         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
28727         (vect_use_loop_mask_for_alignment_p): New function.
28728         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
28729         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
28730         niters_skip argument.  Make sure that the first niters_skip elements
28731         of the first iteration are inactive.
28732         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
28733         Update call to vect_set_loop_masks_directly.
28734         (get_misalign_in_elems): New function, split out from...
28735         (vect_gen_prolog_loop_niters): ...here.
28736         (vect_update_init_of_dr): Take a code argument that specifies whether
28737         the adjustment should be added or subtracted.
28738         (vect_update_init_of_drs): Likewise.
28739         (vect_prepare_for_masked_peels): New function.
28740         (vect_do_peeling): Skip prologue peeling if we're using a mask
28741         instead.  Update call to vect_update_inits_of_drs.
28742         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
28743         mask_skip_niters.
28744         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
28745         alignment.  Do not include the number of peeled iterations in
28746         the minimum threshold in that case.
28747         (vectorizable_induction): Adjust the start value down by
28748         LOOP_VINFO_MASK_SKIP_NITERS iterations.
28749         (vect_transform_loop): Call vect_prepare_for_masked_peels.
28750         Take the number of skipped iterations into account when calculating
28751         the loop bounds.
28752         * tree-vect-stmts.c (vect_gen_while_not): New function.
28754 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28755             Alan Hayward  <alan.hayward@arm.com>
28756             David Sherwood  <david.sherwood@arm.com>
28758         * doc/sourcebuild.texi (vect_fully_masked): Document.
28759         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
28760         default value to 0.
28761         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
28762         split out from...
28763         (vect_analyze_loop_2): ...here. Don't check the vectorization
28764         factor against the number of loop iterations if the loop is
28765         fully-masked.
28767 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28768             Alan Hayward  <alan.hayward@arm.com>
28769             David Sherwood  <david.sherwood@arm.com>
28771         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
28772         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
28773         (dump_groups): Update accordingly.
28774         (iv_use::mem_type): New member variable.
28775         (address_p): New function.
28776         (record_use): Add a mem_type argument and initialize the new
28777         mem_type field.
28778         (record_group_use): Add a mem_type argument.  Use address_p.
28779         Remove obsolete null checks of base_object.  Update call to record_use.
28780         (find_interesting_uses_op): Update call to record_group_use.
28781         (find_interesting_uses_cond): Likewise.
28782         (find_interesting_uses_address): Likewise.
28783         (get_mem_type_for_internal_fn): New function.
28784         (find_address_like_use): Likewise.
28785         (find_interesting_uses_stmt): Try find_address_like_use before
28786         calling find_interesting_uses_op.
28787         (addr_offset_valid_p): Use the iv mem_type field as the type
28788         of the addressed memory.
28789         (add_autoinc_candidates): Likewise.
28790         (get_address_cost): Likewise.
28791         (split_small_address_groups_p): Use address_p.
28792         (split_address_groups): Likewise.
28793         (add_iv_candidate_for_use): Likewise.
28794         (autoinc_possible_for_pair): Likewise.
28795         (rewrite_groups): Likewise.
28796         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
28797         (determine_group_iv_cost): Update after split of USE_ADDRESS.
28798         (get_alias_ptr_type_for_ptr_address): New function.
28799         (rewrite_use_address): Rewrite address uses in calls that were
28800         identified by find_address_like_use.
28802 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28803             Alan Hayward  <alan.hayward@arm.com>
28804             David Sherwood  <david.sherwood@arm.com>
28806         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
28807         TARGET_MEM_REFs.
28808         * gimple-expr.h (is_gimple_addressable: Likewise.
28809         * gimple-expr.c (is_gimple_address): Likewise.
28810         * internal-fn.c (expand_call_mem_ref): New function.
28811         (expand_mask_load_optab_fn): Use it.
28812         (expand_mask_store_optab_fn): Likewise.
28814 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28815             Alan Hayward  <alan.hayward@arm.com>
28816             David Sherwood  <david.sherwood@arm.com>
28818         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
28819         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
28820         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
28821         (cond_umax@var{mode}): Document.
28822         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
28823         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
28824         (cond_umin_optab, cond_umax_optab): New optabs.
28825         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
28826         (COND_IOR, COND_XOR): New internal functions.
28827         * internal-fn.h (get_conditional_internal_fn): Declare.
28828         * internal-fn.c (cond_binary_direct): New macro.
28829         (expand_cond_binary_optab_fn): Likewise.
28830         (direct_cond_binary_optab_supported_p): Likewise.
28831         (get_conditional_internal_fn): New function.
28832         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
28833         Cope with reduction statements that are vectorized as calls rather
28834         than assignments.
28835         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
28836         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
28837         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
28838         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
28839         (UNSPEC_COND_EOR): New unspecs.
28840         (optab): Add mappings for them.
28841         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
28842         (sve_int_op, sve_fp_op): New int attributes.
28844 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28845             Alan Hayward  <alan.hayward@arm.com>
28846             David Sherwood  <david.sherwood@arm.com>
28848         * optabs.def (while_ult_optab): New optab.
28849         * doc/md.texi (while_ult@var{m}@var{n}): Document.
28850         * internal-fn.def (WHILE_ULT): New internal function.
28851         * internal-fn.h (direct_internal_fn_supported_p): New override
28852         that takes two types as argument.
28853         * internal-fn.c (while_direct): New macro.
28854         (expand_while_optab_fn): New function.
28855         (convert_optab_supported_p): Likewise.
28856         (direct_while_optab_supported_p): New macro.
28857         * wide-int.h (wi::udiv_ceil): New function.
28858         * tree-vectorizer.h (rgroup_masks): New structure.
28859         (vec_loop_masks): New typedef.
28860         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
28861         and fully_masked_p.
28862         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
28863         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
28864         (vect_max_vf): New function.
28865         (slpeel_make_loop_iterate_ntimes): Delete.
28866         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
28867         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
28868         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
28869         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
28870         internal-fn.h, stor-layout.h and optabs-query.h.
28871         (vect_set_loop_mask): New function.
28872         (add_preheader_seq): Likewise.
28873         (add_header_seq): Likewise.
28874         (interleave_supported_p): Likewise.
28875         (vect_maybe_permute_loop_masks): Likewise.
28876         (vect_set_loop_masks_directly): Likewise.
28877         (vect_set_loop_condition_masked): Likewise.
28878         (vect_set_loop_condition_unmasked): New function, split out from
28879         slpeel_make_loop_iterate_ntimes.
28880         (slpeel_make_loop_iterate_ntimes): Rename to..
28881         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
28882         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
28883         (vect_do_peeling): Update call accordingly.
28884         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
28885         loops.
28886         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
28887         mask_compare_type, can_fully_mask_p and fully_masked_p.
28888         (release_vec_loop_masks): New function.
28889         (_loop_vec_info): Use it to free the loop masks.
28890         (can_produce_all_loop_masks_p): New function.
28891         (vect_get_max_nscalars_per_iter): Likewise.
28892         (vect_verify_full_masking): Likewise.
28893         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
28894         retries, and free the mask rgroups before retrying.  Check loop-wide
28895         reasons for disallowing fully-masked loops.  Make the final decision
28896         about whether use a fully-masked loop or not.
28897         (vect_estimate_min_profitable_iters): Do not assume that peeling
28898         for the number of iterations will be needed for fully-masked loops.
28899         (vectorizable_reduction): Disable fully-masked loops.
28900         (vectorizable_live_operation): Likewise.
28901         (vect_halve_mask_nunits): New function.
28902         (vect_double_mask_nunits): Likewise.
28903         (vect_record_loop_mask): Likewise.
28904         (vect_get_loop_mask): Likewise.
28905         (vect_transform_loop): Handle the case in which the final loop
28906         iteration might handle a partial vector.  Call vect_set_loop_condition
28907         instead of slpeel_make_loop_iterate_ntimes.
28908         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
28909         (check_load_store_masking): New function.
28910         (prepare_load_store_mask): Likewise.
28911         (vectorizable_store): Handle fully-masked loops.
28912         (vectorizable_load): Likewise.
28913         (supportable_widening_operation): Use vect_halve_mask_nunits for
28914         booleans.
28915         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
28916         (vect_gen_while): New function.
28917         * config/aarch64/aarch64.md (umax<mode>3): New expander.
28918         (aarch64_uqdec<mode>): New insn.
28920 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28921             Alan Hayward  <alan.hayward@arm.com>
28922             David Sherwood  <david.sherwood@arm.com>
28924         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
28925         (reduc_xor_scal_optab): New optabs.
28926         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
28927         (reduc_xor_scal_@var{m}): Document.
28928         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
28929         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
28930         internal functions.
28931         * fold-const-call.c (fold_const_call): Handle them.
28932         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
28933         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
28934         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
28935         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
28936         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
28937         (UNSPEC_XORV): New unspecs.
28938         (optab): Add entries for them.
28939         (BITWISEV): New int iterator.
28940         (bit_reduc_op): New int attributes.
28942 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28943             Alan Hayward  <alan.hayward@arm.com>
28944             David Sherwood  <david.sherwood@arm.com>
28946         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
28947         * internal-fn.def (VEC_SHL_INSERT): New internal function.
28948         * optabs.def (vec_shl_insert_optab): New optab.
28949         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
28950         (duplicate_and_interleave): Likewise.
28951         * tree-vect-loop.c: Include internal-fn.h.
28952         (neutral_op_for_slp_reduction): New function, split out from
28953         get_initial_defs_for_reduction.
28954         (get_initial_def_for_reduction): Handle option 2 for variable-length
28955         vectors by loading the neutral value into a vector and then shifting
28956         the initial value into element 0.
28957         (get_initial_defs_for_reduction): Replace the code argument with
28958         the neutral value calculated by neutral_op_for_slp_reduction.
28959         Use gimple_build_vector for constant-length vectors.
28960         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
28961         but the first group_size elements have a neutral value.
28962         Use duplicate_and_interleave otherwise.
28963         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
28964         Update call to get_initial_defs_for_reduction.  Handle SLP
28965         reductions for variable-length vectors by creating one vector
28966         result for each scalar result, with the elements associated
28967         with other scalar results stubbed out with the neutral value.
28968         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
28969         Require IFN_VEC_SHL_INSERT for double reductions on
28970         variable-length vectors, or SLP reductions that have
28971         a neutral value.  Require can_duplicate_and_interleave_p
28972         support for variable-length unchained SLP reductions if there
28973         is no neutral value, such as for MIN/MAX reductions.  Also require
28974         the number of vector elements to be a multiple of the number of
28975         SLP statements when doing variable-length unchained SLP reductions.
28976         Update call to vect_create_epilog_for_reduction.
28977         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
28978         and remove initial values.
28979         (duplicate_and_interleave): Make public.
28980         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
28981         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
28983 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28984             Alan Hayward  <alan.hayward@arm.com>
28985             David Sherwood  <david.sherwood@arm.com>
28987         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
28988         (can_duplicate_and_interleave_p): New function.
28989         (vect_get_and_check_slp_defs): Take the vector of statements
28990         rather than just the current one.  Remove excess parentheses.
28991         Restriction rejectinon of vect_constant_def and vect_external_def
28992         for variable-length vectors to boolean types, or types for which
28993         can_duplicate_and_interleave_p is false.
28994         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
28995         (duplicate_and_interleave): New function.
28996         (vect_get_constant_vectors): Use gimple_build_vector for
28997         constant-length vectors and suitable variable-length constant
28998         vectors.  Use duplicate_and_interleave for other variable-length
28999         vectors.  Don't defer the update when inserting new statements.
29001 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29002             Alan Hayward  <alan.hayward@arm.com>
29003             David Sherwood  <david.sherwood@arm.com>
29005         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
29006         min_profitable_iters doesn't go negative.
29008 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29009             Alan Hayward  <alan.hayward@arm.com>
29010             David Sherwood  <david.sherwood@arm.com>
29012         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
29013         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
29014         * optabs.def (vec_mask_load_lanes_optab): New optab.
29015         (vec_mask_store_lanes_optab): Likewise.
29016         * internal-fn.def (MASK_LOAD_LANES): New internal function.
29017         (MASK_STORE_LANES): Likewise.
29018         * internal-fn.c (mask_load_lanes_direct): New macro.
29019         (mask_store_lanes_direct): Likewise.
29020         (expand_mask_load_optab_fn): Handle masked operations.
29021         (expand_mask_load_lanes_optab_fn): New macro.
29022         (expand_mask_store_optab_fn): Handle masked operations.
29023         (expand_mask_store_lanes_optab_fn): New macro.
29024         (direct_mask_load_lanes_optab_supported_p): Likewise.
29025         (direct_mask_store_lanes_optab_supported_p): Likewise.
29026         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
29027         parameter.
29028         (vect_load_lanes_supported): Likewise.
29029         * tree-vect-data-refs.c (strip_conversion): New function.
29030         (can_group_stmts_p): Likewise.
29031         (vect_analyze_data_ref_accesses): Use it instead of checking
29032         for a pair of assignments.
29033         (vect_store_lanes_supported): Take a masked_p parameter.
29034         (vect_load_lanes_supported): Likewise.
29035         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
29036         vect_store_lanes_supported and vect_load_lanes_supported.
29037         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
29038         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
29039         parameter.  Don't allow gaps for masked accesses.
29040         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
29041         and vect_load_lanes_supported.
29042         (get_load_store_type): Take a masked_p parameter and update
29043         call to get_group_load_store_type.
29044         (vectorizable_store): Update call to get_load_store_type.
29045         Handle IFN_MASK_STORE_LANES.
29046         (vectorizable_load): Update call to get_load_store_type.
29047         Handle IFN_MASK_LOAD_LANES.
29049 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29050             Alan Hayward  <alan.hayward@arm.com>
29051             David Sherwood  <david.sherwood@arm.com>
29053         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
29054         modes for SVE.
29055         * config/aarch64/aarch64-protos.h
29056         (aarch64_sve_struct_memory_operand_p): Declare.
29057         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
29058         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
29059         (VPRED, vpred): Handle SVE structure modes.
29060         * config/aarch64/constraints.md (Utx): New constraint.
29061         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
29062         (aarch64_sve_struct_nonimmediate_operand): New predicates.
29063         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
29064         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
29065         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
29066         structure modes.  Split into pieces after RA.
29067         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
29068         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
29069         New patterns.
29070         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
29071         SVE structure modes.
29072         (aarch64_classify_address): Likewise.
29073         (sizetochar): Move earlier in file.
29074         (aarch64_print_operand): Handle SVE register lists.
29075         (aarch64_array_mode): New function.
29076         (aarch64_sve_struct_memory_operand_p): Likewise.
29077         (TARGET_ARRAY_MODE): Redefine.
29079 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29080             Alan Hayward  <alan.hayward@arm.com>
29081             David Sherwood  <david.sherwood@arm.com>
29083         * target.def (array_mode): New target hook.
29084         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
29085         * doc/tm.texi: Regenerate.
29086         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
29087         * hooks.c (hook_optmode_mode_uhwi_none): New function.
29088         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
29089         targetm.array_mode.
29090         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
29091         type sizes.
29093 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29094             Alan Hayward  <alan.hayward@arm.com>
29095             David Sherwood  <david.sherwood@arm.com>
29097         * fold-const.c (fold_binary_loc): Check the argument types
29098         rather than the result type when testing for a vector operation.
29100 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29102         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
29103         * doc/tm.texi: Regenerate.
29105 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29106             Alan Hayward  <alan.hayward@arm.com>
29107             David Sherwood  <david.sherwood@arm.com>
29109         * doc/invoke.texi (-msve-vector-bits=): Document new option.
29110         (sve): Document new AArch64 extension.
29111         * doc/md.texi (w): Extend the description of the AArch64
29112         constraint to include SVE vectors.
29113         (Upl, Upa): Document new AArch64 predicate constraints.
29114         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
29115         enum.
29116         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
29117         (msve-vector-bits=): New option.
29118         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
29119         SVE when these are disabled.
29120         (sve): New extension.
29121         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
29122         modes.  Adjust their number of units based on aarch64_sve_vg.
29123         (MAX_BITSIZE_MODE_ANY_MODE): Define.
29124         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
29125         aarch64_addr_query_type.
29126         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
29127         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
29128         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
29129         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
29130         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
29131         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
29132         (aarch64_simd_imm_zero_p): Delete.
29133         (aarch64_check_zero_based_sve_index_immediate): Declare.
29134         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
29135         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
29136         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
29137         (aarch64_sve_float_mul_immediate_p): Likewise.
29138         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
29139         rather than an rtx.
29140         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
29141         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
29142         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
29143         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
29144         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
29145         (aarch64_regmode_natural_size): Likewise.
29146         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
29147         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
29148         left one place.
29149         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
29150         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
29151         for VG and the SVE predicate registers.
29152         (V_ALIASES): Add a "z"-prefixed alias.
29153         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
29154         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
29155         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
29156         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
29157         (REG_CLASS_NAMES): Add entries for them.
29158         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
29159         and the predicate registers.
29160         (aarch64_sve_vg): Declare.
29161         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
29162         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
29163         (REGMODE_NATURAL_SIZE): Define.
29164         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
29165         SVE macros.
29166         * config/aarch64/aarch64.c: Include cfgrtl.h.
29167         (simd_immediate_info): Add a constructor for series vectors,
29168         and an associated step field.
29169         (aarch64_sve_vg): New variable.
29170         (aarch64_dbx_register_number): Handle VG and the predicate registers.
29171         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
29172         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
29173         (VEC_ANY_DATA, VEC_STRUCT): New constants.
29174         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
29175         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
29176         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
29177         (aarch64_get_mask_mode): New functions.
29178         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
29179         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
29180         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
29181         predicate modes and predicate registers.  Explicitly restrict
29182         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
29183         to store a vector mode if it is recognized by
29184         aarch64_classify_vector_mode.
29185         (aarch64_regmode_natural_size): New function.
29186         (aarch64_hard_regno_caller_save_mode): Return the original mode
29187         for predicates.
29188         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
29189         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
29190         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
29191         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
29192         functions.
29193         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
29194         does not overlap dest if the function is frame-related.  Handle
29195         SVE constants.
29196         (aarch64_split_add_offset): New function.
29197         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
29198         them aarch64_add_offset.
29199         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
29200         and update call to aarch64_sub_sp.
29201         (aarch64_add_cfa_expression): New function.
29202         (aarch64_expand_prologue): Pass extra temporary registers to the
29203         functions above.  Handle the case in which we need to emit new
29204         DW_CFA_expressions for registers that were originally saved
29205         relative to the stack pointer, but now have to be expressed
29206         relative to the frame pointer.
29207         (aarch64_output_mi_thunk): Pass extra temporary registers to the
29208         functions above.
29209         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
29210         IP0 and IP1 values for SVE frames.
29211         (aarch64_expand_vec_series): New function.
29212         (aarch64_expand_sve_widened_duplicate): Likewise.
29213         (aarch64_expand_sve_const_vector): Likewise.
29214         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
29215         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
29216         into the register, rather than emitting a SET directly.
29217         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
29218         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
29219         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
29220         (offset_9bit_signed_scaled_p): New functions.
29221         (aarch64_replicate_bitmask_imm): New function.
29222         (aarch64_bitmask_imm): Use it.
29223         (aarch64_cannot_force_const_mem): Reject expressions involving
29224         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
29225         (aarch64_classify_index): Handle SVE indices, by requiring
29226         a plain register index with a scale that matches the element size.
29227         (aarch64_classify_address): Handle SVE addresses.  Assert that
29228         the mode of the address is VOIDmode or an integer mode.
29229         Update call to aarch64_classify_symbol.
29230         (aarch64_classify_symbolic_expression): Update call to
29231         aarch64_classify_symbol.
29232         (aarch64_const_vec_all_in_range_p): New function.
29233         (aarch64_print_vector_float_operand): Likewise.
29234         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
29235         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
29236         and the FP immediates 1.0 and 0.5.
29237         (aarch64_print_address_internal): Handle SVE addresses.
29238         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
29239         (aarch64_regno_regclass): Handle predicate registers.
29240         (aarch64_secondary_reload): Handle big-endian reloads of SVE
29241         data modes.
29242         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
29243         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
29244         (aarch64_convert_sve_vector_bits): New function.
29245         (aarch64_override_options): Use it to handle -msve-vector-bits=.
29246         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
29247         rather than an rtx.
29248         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
29249         Handle SVE vector and predicate modes.  Accept VL-based constants
29250         that need only one temporary register, and VL offsets that require
29251         no temporary registers.
29252         (aarch64_conditional_register_usage): Mark the predicate registers
29253         as fixed if SVE isn't available.
29254         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
29255         Return true for SVE vector and predicate modes.
29256         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
29257         rather than an unsigned int.  Handle SVE modes.
29258         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
29259         SVE modes.
29260         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
29261         if SVE is enabled.
29262         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
29263         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
29264         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
29265         (aarch64_sve_float_mul_immediate_p): New functions.
29266         (aarch64_sve_valid_immediate): New function.
29267         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
29268         Explicitly reject structure modes.  Check for INDEX constants.
29269         Handle PTRUE and PFALSE constants.
29270         (aarch64_check_zero_based_sve_index_immediate): New function.
29271         (aarch64_simd_imm_zero_p): Delete.
29272         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
29273         vector modes.  Accept constants in the range of CNT[BHWD].
29274         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
29275         ask for an Advanced SIMD mode.
29276         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
29277         (aarch64_simd_vector_alignment): Handle SVE predicates.
29278         (aarch64_vectorize_preferred_vector_alignment): New function.
29279         (aarch64_simd_vector_alignment_reachable): Use it instead of
29280         the vector size.
29281         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
29282         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
29283         functions.
29284         (MAX_VECT_LEN): Delete.
29285         (expand_vec_perm_d): Add a vec_flags field.
29286         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
29287         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
29288         (aarch64_evpc_ext): Don't apply a big-endian lane correction
29289         for SVE modes.
29290         (aarch64_evpc_rev): Rename to...
29291         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
29292         (aarch64_evpc_rev_global): New function.
29293         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
29294         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
29295         MAX_VECT_LEN.
29296         (aarch64_evpc_sve_tbl): New function.
29297         (aarch64_expand_vec_perm_const_1): Update after rename of
29298         aarch64_evpc_rev.  Handle SVE permutes too, trying
29299         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
29300         than aarch64_evpc_tbl.
29301         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
29302         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
29303         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
29304         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
29305         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
29306         (aarch64_expand_sve_vcond): New functions.
29307         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
29308         of aarch64_vector_mode_p.
29309         (aarch64_dwarf_poly_indeterminate_value): New function.
29310         (aarch64_compute_pressure_classes): Likewise.
29311         (aarch64_can_change_mode_class): Likewise.
29312         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
29313         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
29314         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
29315         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
29316         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
29317         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
29318         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
29319         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
29320         constraints.
29321         (Dn, Dl, Dr): Accept const as well as const_vector.
29322         (Dz): Likewise.  Compare against CONST0_RTX.
29323         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
29324         of "vector" where appropriate.
29325         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
29326         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
29327         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
29328         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
29329         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
29330         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
29331         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
29332         (v_int_equiv): Extend to SVE modes.
29333         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
29334         mode attributes.
29335         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
29336         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
29337         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
29338         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
29339         (SVE_COND_FP_CMP): New int iterators.
29340         (perm_hilo): Handle the new unpack unspecs.
29341         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
29342         attributes.
29343         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
29344         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
29345         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
29346         (aarch64_equality_operator, aarch64_constant_vector_operand)
29347         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
29348         (aarch64_sve_nonimmediate_operand): Likewise.
29349         (aarch64_sve_general_operand): Likewise.
29350         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
29351         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
29352         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
29353         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
29354         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
29355         (aarch64_sve_float_arith_immediate): Likewise.
29356         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
29357         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
29358         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
29359         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
29360         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
29361         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
29362         (aarch64_sve_float_arith_operand): Likewise.
29363         (aarch64_sve_float_arith_with_sub_operand): Likewise.
29364         (aarch64_sve_float_mul_operand): Likewise.
29365         (aarch64_sve_vec_perm_operand): Likewise.
29366         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
29367         (aarch64_mov_operand): Accept const_poly_int and const_vector.
29368         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
29369         as well as const_vector.
29370         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
29371         in file.  Use CONST0_RTX and CONSTM1_RTX.
29372         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
29373         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
29374         Use aarch64_simd_imm_zero.
29375         * config/aarch64/aarch64-sve.md: New file.
29376         * config/aarch64/aarch64.md: Include it.
29377         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
29378         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
29379         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
29380         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
29381         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
29382         (sve): New attribute.
29383         (enabled): Disable instructions with the sve attribute unless
29384         TARGET_SVE.
29385         (movqi, movhi): Pass CONST_POLY_INT operaneds through
29386         aarch64_expand_mov_immediate.
29387         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
29388         CNT[BHSD] immediates.
29389         (movti): Split CONST_POLY_INT moves into two halves.
29390         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
29391         Split additions that need a temporary here if the destination
29392         is the stack pointer.
29393         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
29394         (*add<mode>3_poly_1): New instruction.
29395         (set_clobber_cc): New expander.
29397 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29399         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
29400         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
29401         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
29402         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
29403         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
29404         Change innermode from fixed_mode_size to machine_mode.
29405         (simplify_subreg): Update call accordingly.  Handle a constant-sized
29406         subreg of a variable-length CONST_VECTOR.
29408 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29409             Alan Hayward  <alan.hayward@arm.com>
29410             David Sherwood  <david.sherwood@arm.com>
29412         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
29413         (add_offset_to_base): New function, split out from...
29414         (create_mem_ref): ...here.  When handling a scale other than 1,
29415         check first whether the address is valid without the offset.
29416         Add it into the base if so, leaving the index and scale as-is.
29418 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
29420         PR c++/83778
29421         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
29422         fold_for_warn before checking if arg2 is INTEGER_CST.
29424 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
29426         * config/rs6000/predicates.md (load_multiple_operation): Delete.
29427         (store_multiple_operation): Delete.
29428         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
29429         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
29430         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
29431         guarded by TARGET_STRING.
29432         (rs6000_output_load_multiple): Delete.
29433         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
29434         OPTION_MASK_STRING / TARGET_STRING handling.
29435         (print_operand) <'N', 'O'>: Add comment that these are unused now.
29436         (const rs6000_opt_masks) <"string">: Change mask to 0.
29437         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
29438         (MASK_STRING): Delete.
29439         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
29440         parts.  Simplify.
29441         (load_multiple): Delete.
29442         (*ldmsi8): Delete.
29443         (*ldmsi7): Delete.
29444         (*ldmsi6): Delete.
29445         (*ldmsi5): Delete.
29446         (*ldmsi4): Delete.
29447         (*ldmsi3): Delete.
29448         (store_multiple): Delete.
29449         (*stmsi8): Delete.
29450         (*stmsi7): Delete.
29451         (*stmsi6): Delete.
29452         (*stmsi5): Delete.
29453         (*stmsi4): Delete.
29454         (*stmsi3): Delete.
29455         (movmemsi_8reg): Delete.
29456         (corresponding unnamed define_insn): Delete.
29457         (movmemsi_6reg): Delete.
29458         (corresponding unnamed define_insn): Delete.
29459         (movmemsi_4reg): Delete.
29460         (corresponding unnamed define_insn): Delete.
29461         (movmemsi_2reg): Delete.
29462         (corresponding unnamed define_insn): Delete.
29463         (movmemsi_1reg): Delete.
29464         (corresponding unnamed define_insn): Delete.
29465         * config/rs6000/rs6000.opt (mno-string): New.
29466         (mstring): Replace by deprecation warning stub.
29467         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
29469 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
29471         * regrename.c (regrename_do_replace): If replacing the same
29472         reg multiple times, try to reuse last created gen_raw_REG.
29474         PR debug/81155
29475         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
29476         main to workaround a bug in GDB.
29478 2018-01-12  Tom de Vries  <tom@codesourcery.com>
29480         PR target/83737
29481         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
29483 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
29485         PR rtl-optimization/80481
29486         * ira-color.c (get_cap_member): New function.
29487         (allocnos_conflict_by_live_ranges_p): Use it.
29488         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
29489         (setup_slot_coalesced_allocno_live_ranges): Ditto.
29491 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
29493         PR target/83628
29494         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
29495         (*saddl_se_1): Ditto.
29496         (*ssubsi_1): Ditto.
29497         (*ssubl_se_1): Ditto.
29499 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29501         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
29502         rather than wi::to_widest for DR_INITs.
29503         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
29504         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
29505         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
29506         INTEGER_CSTs.
29507         (vect_analyze_group_access_1): Note that here.
29509 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29511         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
29512         polynomial type sizes.
29514 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29516         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
29517         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
29518         (gimple_add_tmp_var): Likewise.
29520 2018-01-12  Martin Liska  <mliska@suse.cz>
29522         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
29523         (gimple_alloc_sizes): Likewise.
29524         (dump_gimple_statistics): Use PRIu64 in printf format.
29525         * gimple.h: Change uint64_t to int.
29527 2018-01-12  Martin Liska  <mliska@suse.cz>
29529         * tree-core.h: Use uint64_t instead of int.
29530         * tree.c (tree_node_counts): Likewise.
29531         (tree_node_sizes): Likewise.
29532         (dump_tree_statistics): Use PRIu64 in printf format.
29534 2018-01-12  Martin Liska  <mliska@suse.cz>
29536         * Makefile.in: As qsort_chk is implemented in vec.c, add
29537         vec.o to linkage of gencfn-macros.
29538         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
29539         passing the info to record_node_allocation_statistics.
29540         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
29541         and pass the info.
29542         * ggc-common.c (struct ggc_usage): Add operator== and use
29543         it in operator< and compare function.
29544         * mem-stats.h (struct mem_usage): Likewise.
29545         * vec.c (struct vec_usage): Remove operator< and compare
29546         function. Can be simply inherited.
29548 2018-01-12  Martin Jambor  <mjambor@suse.cz>
29550         PR target/81616
29551         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
29552         * tree-ssa-math-opts.c: Include domwalk.h.
29553         (convert_mult_to_fma_1): New function.
29554         (fma_transformation_info): New type.
29555         (fma_deferring_state): Likewise.
29556         (cancel_fma_deferring): New function.
29557         (result_of_phi): Likewise.
29558         (last_fma_candidate_feeds_initial_phi): Likewise.
29559         (convert_mult_to_fma): Added deferring logic, split actual
29560         transformation to convert_mult_to_fma_1.
29561         (math_opts_dom_walker): New type.
29562         (math_opts_dom_walker::after_dom_children): New method, body moved
29563         here from pass_optimize_widening_mul::execute, added deferring logic
29564         bits.
29565         (pass_optimize_widening_mul::execute): Moved most of code to
29566         math_opts_dom_walker::after_dom_children.
29567         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
29568         * config/i386/i386.c (ix86_option_override_internal): Added
29569         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
29571 2018-01-12  Richard Biener  <rguenther@suse.de>
29573         PR debug/83157
29574         * dwarf2out.c (gen_variable_die): Do not reset old_die for
29575         inline instance vars.
29577 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
29579         PR target/81819
29580         * config/rx/rx.c (rx_is_restricted_memory_address):
29581         Handle SUBREG case.
29583 2018-01-12  Richard Biener  <rguenther@suse.de>
29585         PR tree-optimization/80846
29586         * target.def (split_reduction): New target hook.
29587         * targhooks.c (default_split_reduction): New function.
29588         * targhooks.h (default_split_reduction): Declare.
29589         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
29590         target requests first reduce vectors by combining low and high
29591         parts.
29592         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
29593         (get_vectype_for_scalar_type_and_size): Export.
29594         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
29595         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
29596         * doc/tm.texi: Regenerate.
29597         * config/i386/i386.c (ix86_split_reduction): Implement
29598         TARGET_VECTORIZE_SPLIT_REDUCTION.
29600 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
29602         PR target/83368
29603         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
29604         in PIC mode except for TARGET_VXWORKS_RTP.
29605         * config/sparc/sparc.c: Include cfgrtl.h.
29606         (TARGET_INIT_PIC_REG): Define.
29607         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
29608         (sparc_pic_register_p): New predicate.
29609         (sparc_legitimate_address_p): Use it.
29610         (sparc_legitimize_pic_address): Likewise.
29611         (sparc_delegitimize_address): Likewise.
29612         (sparc_mode_dependent_address_p): Likewise.
29613         (gen_load_pcrel_sym): Remove 4th parameter.
29614         (load_got_register): Adjust call to above.  Remove obsolete stuff.
29615         (sparc_expand_prologue): Do not call load_got_register here.
29616         (sparc_flat_expand_prologue): Likewise.
29617         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
29618         (sparc_use_pseudo_pic_reg): New function.
29619         (sparc_init_pic_reg): Likewise.
29620         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
29621         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
29623 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
29625         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
29626         Add item for branch_cost.
29628 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
29630         PR rtl-optimization/83565
29631         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
29632         not extend the result to a larger mode for rotate operations.
29633         (num_sign_bit_copies1): Likewise.
29635 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
29637         PR target/40411
29638         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
29639         -symbolic.
29640         Use values-Xc.o for -pedantic.
29641         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
29643 2018-01-12  Martin Liska  <mliska@suse.cz>
29645         PR ipa/83054
29646         * ipa-devirt.c (final_warning_record::grow_type_warnings):
29647         New function.
29648         (possible_polymorphic_call_targets): Use it.
29649         (ipa_devirt): Likewise.
29651 2018-01-12  Martin Liska  <mliska@suse.cz>
29653         * profile-count.h (enum profile_quality): Use 0 as invalid
29654         enum value of profile_quality.
29656 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
29658         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
29659         -mext-string options.
29661 2018-01-12  Richard Biener  <rguenther@suse.de>
29663         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
29664         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
29665         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
29666         Likewise.
29667         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
29669 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
29671         * configure.ac (--with-long-double-format): Add support for the
29672         configuration option to change the default long double format on
29673         PowerPC systems.
29674         * config.gcc (powerpc*-linux*-*): Likewise.
29675         * configure: Regenerate.
29676         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
29677         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
29678         used without modification.
29680 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29682         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
29683         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
29684         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
29685         MISC_BUILTIN_SPEC_BARRIER.
29686         (rs6000_init_builtins): Likewise.
29687         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
29688         enum value.
29689         (speculation_barrier): New define_insn.
29690         * doc/extend.texi: Document __builtin_speculation_barrier.
29692 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
29694         PR target/83203
29695         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
29696         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
29697         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
29698         iterators.
29699         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
29700         integral modes instead of "ss" and "sd".
29701         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
29702         vectors with 32-bit and 64-bit elements.
29703         (vecdupssescalarmodesuffix): New mode attribute.
29704         (vec_dup<mode>): Use it.
29706 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
29708         PR target/83330
29709         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
29710         frame if argument is passed on stack.
29712 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
29714         PR target/82682
29715         * ree.c (combine_reaching_defs): Optimize also
29716         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
29717         reg2=any_extend(exp); reg1=reg2;, formatting fix.
29719 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
29721         PR middle-end/83189
29722         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
29724 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
29726         PR middle-end/83718
29727         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
29728         after they are computed.
29730 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
29732         PR tree-optimization/83695
29733         * gimple-loop-linterchange.cc
29734         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
29735         reset cached scev information after interchange.
29736         (pass_linterchange::execute): Remove call to scev_reset_htab.
29738 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29740         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
29741         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
29742         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
29743         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
29744         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
29745         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
29746         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
29747         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
29748         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
29749         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
29750         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
29751         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
29752         (V_lane_reg): Likewise.
29753         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
29754         New define_expand.
29755         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
29756         (vfmal_lane_low<mode>_intrinsic,
29757         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
29758         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
29759         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
29760         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
29761         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
29762         vfmsl_lane_high<mode>_intrinsic): New define_insns.
29764 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29766         * config/arm/arm-cpus.in (fp16fml): New feature.
29767         (ALL_SIMD): Add fp16fml.
29768         (armv8.2-a): Add fp16fml as an option.
29769         (armv8.3-a): Likewise.
29770         (armv8.4-a): Add fp16fml as part of fp16.
29771         * config/arm/arm.h (TARGET_FP16FML): Define.
29772         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
29773         when appropriate.
29774         * config/arm/arm-modes.def (V2HF): Define.
29775         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
29776         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
29777         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
29778         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
29779         vfmsl_low, vfmsl_high): New set of builtins.
29780         * config/arm/iterators.md (PLUSMINUS): New code iterator.
29781         (vfml_op): New code attribute.
29782         (VFMLHALVES): New int iterator.
29783         (VFML, VFMLSEL): New mode attributes.
29784         (V_reg): Define mapping for V2HF.
29785         (V_hi, V_lo): New mode attributes.
29786         (VF_constraint): Likewise.
29787         (vfml_half, vfml_half_selector): New int attributes.
29788         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
29789         define_expand.
29790         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
29791         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
29792         New define_insn.
29793         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
29794         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
29795         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
29796         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
29797         documentation.
29798         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
29799         Document new effective target and option set.
29801 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29803         * config/arm/arm-cpus.in (armv8_4): New feature.
29804         (ARMv8_4a): New fgroup.
29805         (armv8.4-a): New arch.
29806         * config/arm/arm-tables.opt: Regenerate.
29807         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
29808         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
29809         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
29810         Add matching rules for -march=armv8.4-a and extensions.
29811         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
29813 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
29815         PR target/81821
29816         * config/rx/rx.md (BW): New mode attribute.
29817         (sync_lock_test_and_setsi): Add mode suffix to insn output.
29819 2018-01-11  Richard Biener  <rguenther@suse.de>
29821         PR tree-optimization/83435
29822         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
29823         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
29824         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
29826 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29827             Alan Hayward  <alan.hayward@arm.com>
29828             David Sherwood  <david.sherwood@arm.com>
29830         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
29831         field.
29832         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
29833         (aarch64_print_address_internal): Use it to check for a zero offset.
29835 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29836             Alan Hayward  <alan.hayward@arm.com>
29837             David Sherwood  <david.sherwood@arm.com>
29839         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
29840         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
29841         Return a poly_int64 rather than a HOST_WIDE_INT.
29842         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
29843         rather than a HOST_WIDE_INT.
29844         * config/aarch64/aarch64.h (aarch64_frame): Protect with
29845         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
29846         hard_fp_offset, frame_size, initial_adjust, callee_offset and
29847         final_offset from HOST_WIDE_INT to poly_int64.
29848         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
29849         to_constant when getting the number of units in an Advanced SIMD
29850         mode.
29851         (aarch64_builtin_vectorized_function): Check for a constant number
29852         of units.
29853         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
29854         GET_MODE_SIZE.
29855         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
29856         attribute instead of GET_MODE_NUNITS.
29857         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
29858         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
29859         GET_MODE_SIZE for fixed-size registers.
29860         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
29861         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
29862         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
29863         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
29864         (aarch64_print_operand, aarch64_print_address_internal)
29865         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
29866         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
29867         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
29868         Handle polynomial GET_MODE_SIZE.
29869         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
29870         wider than SImode without modification.
29871         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
29872         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
29873         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
29874         passing and returning SVE modes.
29875         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
29876         rather than GEN_INT.
29877         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
29878         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
29879         (aarch64_allocate_and_probe_stack_space): Likewise.
29880         (aarch64_layout_frame): Cope with polynomial offsets.
29881         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
29882         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
29883         polynomial offsets.
29884         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
29885         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
29886         poly_int64 rather than a HOST_WIDE_INT.
29887         (aarch64_get_separate_components, aarch64_process_components)
29888         (aarch64_expand_prologue, aarch64_expand_epilogue)
29889         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
29890         (aarch64_anchor_offset): New function, split out from...
29891         (aarch64_legitimize_address): ...here.
29892         (aarch64_builtin_vectorization_cost): Handle polynomial
29893         TYPE_VECTOR_SUBPARTS.
29894         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
29895         GET_MODE_NUNITS.
29896         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
29897         number of elements from the PARALLEL rather than the mode.
29898         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
29899         rather than GET_MODE_BITSIZE.
29900         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
29901         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
29902         (aarch64_expand_vec_perm_const_1): Handle polynomial
29903         d->perm.length () and d->perm elements.
29904         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
29905         Apply to_constant to d->perm elements.
29906         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
29907         polynomial CONST_VECTOR_NUNITS.
29908         (aarch64_move_pointer): Take amount as a poly_int64 rather
29909         than an int.
29910         (aarch64_progress_pointer): Avoid temporary variable.
29911         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
29912         the mode attribute instead of GET_MODE.
29914 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29915             Alan Hayward  <alan.hayward@arm.com>
29916             David Sherwood  <david.sherwood@arm.com>
29918         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
29919         x exists before using it.
29920         (aarch64_add_constant_internal): Rename to...
29921         (aarch64_add_offset_1): ...this.  Replace regnum with separate
29922         src and dest rtxes.  Handle the case in which they're different,
29923         including when the offset is zero.  Replace scratchreg with an rtx.
29924         Use 2 additions if there is no spare register into which we can
29925         move a 16-bit constant.
29926         (aarch64_add_constant): Delete.
29927         (aarch64_add_offset): Replace reg with separate src and dest
29928         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
29929         Use aarch64_add_offset_1.
29930         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
29931         an rtx rather than an int.  Take the delta as a poly_int64
29932         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
29933         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
29934         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
29935         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
29936         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
29937         and aarch64_add_sp.
29938         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
29939         aarch64_add_constant.
29941 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29943         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
29944         Use scalar_float_mode.
29946 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29948         * config/aarch64/aarch64-simd.md
29949         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
29950         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
29951         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
29952         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
29953         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
29954         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
29955         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
29956         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
29957         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
29958         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
29960 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
29962         PR target/83514
29963         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
29964         targ_options->x_arm_arch_string is non NULL.
29966 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
29968         * config/aarch64/aarch64.h
29969         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
29971 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
29973         PR target/82096
29974         * expmed.c (emit_store_flag_force): Swap if const op0
29975         and change VOIDmode to mode of op0.
29977 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29979         PR rtl-optimization/83761
29980         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
29981         than bytes to mode_for_size.
29983 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
29985         PR middle-end/83189
29986         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
29987         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
29988         profile.
29990 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
29992         PR middle-end/83575
29993         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
29994         when in layout mode.
29995         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
29996         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
29997         partition fixup.
29999 2018-01-10  Michael Collison  <michael.collison@arm.com>
30001         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
30002         * config/aarch64/aarch64-option-extension.def: Add
30003         AARCH64_OPT_EXTENSION of 'fp16fml'.
30004         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30005         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
30006         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
30007         * config/aarch64/constraints.md (Ui7): New constraint.
30008         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
30009         (VFMLA_SEL_W): Ditto.
30010         (f16quad): Ditto.
30011         (f16mac1): Ditto.
30012         (VFMLA16_LOW): New int iterator.
30013         (VFMLA16_HIGH): Ditto.
30014         (UNSPEC_FMLAL): New unspec.
30015         (UNSPEC_FMLSL): Ditto.
30016         (UNSPEC_FMLAL2): Ditto.
30017         (UNSPEC_FMLSL2): Ditto.
30018         (f16mac): New code attribute.
30019         * config/aarch64/aarch64-simd-builtins.def
30020         (aarch64_fmlal_lowv2sf): Ditto.
30021         (aarch64_fmlsl_lowv2sf): Ditto.
30022         (aarch64_fmlalq_lowv4sf): Ditto.
30023         (aarch64_fmlslq_lowv4sf): Ditto.
30024         (aarch64_fmlal_highv2sf): Ditto.
30025         (aarch64_fmlsl_highv2sf): Ditto.
30026         (aarch64_fmlalq_highv4sf): Ditto.
30027         (aarch64_fmlslq_highv4sf): Ditto.
30028         (aarch64_fmlal_lane_lowv2sf): Ditto.
30029         (aarch64_fmlsl_lane_lowv2sf): Ditto.
30030         (aarch64_fmlal_laneq_lowv2sf): Ditto.
30031         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
30032         (aarch64_fmlalq_lane_lowv4sf): Ditto.
30033         (aarch64_fmlsl_lane_lowv4sf): Ditto.
30034         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
30035         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
30036         (aarch64_fmlal_lane_highv2sf): Ditto.
30037         (aarch64_fmlsl_lane_highv2sf): Ditto.
30038         (aarch64_fmlal_laneq_highv2sf): Ditto.
30039         (aarch64_fmlsl_laneq_highv2sf): Ditto.
30040         (aarch64_fmlalq_lane_highv4sf): Ditto.
30041         (aarch64_fmlsl_lane_highv4sf): Ditto.
30042         (aarch64_fmlalq_laneq_highv4sf): Ditto.
30043         (aarch64_fmlsl_laneq_highv4sf): Ditto.
30044         * config/aarch64/aarch64-simd.md:
30045         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
30046         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
30047         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
30048         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
30049         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
30050         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
30051         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
30052         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
30053         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
30054         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
30055         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
30056         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
30057         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
30058         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
30059         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
30060         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
30061         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
30062         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
30063         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
30064         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
30065         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
30066         (vfmlsl_low_u32): Ditto.
30067         (vfmlalq_low_u32): Ditto.
30068         (vfmlslq_low_u32): Ditto.
30069         (vfmlal_high_u32): Ditto.
30070         (vfmlsl_high_u32): Ditto.
30071         (vfmlalq_high_u32): Ditto.
30072         (vfmlslq_high_u32): Ditto.
30073         (vfmlal_lane_low_u32): Ditto.
30074         (vfmlsl_lane_low_u32): Ditto.
30075         (vfmlal_laneq_low_u32): Ditto.
30076         (vfmlsl_laneq_low_u32): Ditto.
30077         (vfmlalq_lane_low_u32): Ditto.
30078         (vfmlslq_lane_low_u32): Ditto.
30079         (vfmlalq_laneq_low_u32): Ditto.
30080         (vfmlslq_laneq_low_u32): Ditto.
30081         (vfmlal_lane_high_u32): Ditto.
30082         (vfmlsl_lane_high_u32): Ditto.
30083         (vfmlal_laneq_high_u32): Ditto.
30084         (vfmlsl_laneq_high_u32): Ditto.
30085         (vfmlalq_lane_high_u32): Ditto.
30086         (vfmlslq_lane_high_u32): Ditto.
30087         (vfmlalq_laneq_high_u32): Ditto.
30088         (vfmlslq_laneq_high_u32): Ditto.
30089         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
30090         (AARCH64_FL_FOR_ARCH8_4): New.
30091         (AARCH64_ISA_F16FML): New ISA flag.
30092         (TARGET_F16FML): New feature flag for fp16fml.
30093         (doc/invoke.texi): Document new fp16fml option.
30095 2018-01-10  Michael Collison  <michael.collison@arm.com>
30097         * config/aarch64/aarch64-builtins.c:
30098         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
30099         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30100         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
30101         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
30102         (AARCH64_ISA_SHA3): New ISA flag.
30103         (TARGET_SHA3): New feature flag for sha3.
30104         * config/aarch64/iterators.md (sha512_op): New int attribute.
30105         (CRYPTO_SHA512): New int iterator.
30106         (UNSPEC_SHA512H): New unspec.
30107         (UNSPEC_SHA512H2): Ditto.
30108         (UNSPEC_SHA512SU0): Ditto.
30109         (UNSPEC_SHA512SU1): Ditto.
30110         * config/aarch64/aarch64-simd-builtins.def
30111         (aarch64_crypto_sha512hqv2di): New builtin.
30112         (aarch64_crypto_sha512h2qv2di): Ditto.
30113         (aarch64_crypto_sha512su0qv2di): Ditto.
30114         (aarch64_crypto_sha512su1qv2di): Ditto.
30115         (aarch64_eor3qv8hi): Ditto.
30116         (aarch64_rax1qv2di): Ditto.
30117         (aarch64_xarqv2di): Ditto.
30118         (aarch64_bcaxqv8hi): Ditto.
30119         * config/aarch64/aarch64-simd.md:
30120         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
30121         (aarch64_crypto_sha512su0qv2di): Ditto.
30122         (aarch64_crypto_sha512su1qv2di): Ditto.
30123         (aarch64_eor3qv8hi): Ditto.
30124         (aarch64_rax1qv2di): Ditto.
30125         (aarch64_xarqv2di): Ditto.
30126         (aarch64_bcaxqv8hi): Ditto.
30127         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
30128         (vsha512h2q_u64): Ditto.
30129         (vsha512su0q_u64): Ditto.
30130         (vsha512su1q_u64): Ditto.
30131         (veor3q_u16): Ditto.
30132         (vrax1q_u64): Ditto.
30133         (vxarq_u64): Ditto.
30134         (vbcaxq_u16): Ditto.
30135         * config/arm/types.md (crypto_sha512): New type attribute.
30136         (crypto_sha3): Ditto.
30137         (doc/invoke.texi): Document new sha3 option.
30139 2018-01-10  Michael Collison  <michael.collison@arm.com>
30141         * config/aarch64/aarch64-builtins.c:
30142         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
30143         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30144         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
30145         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
30146         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
30147         (AARCH64_ISA_SM4): New ISA flag.
30148         (TARGET_SM4): New feature flag for sm4.
30149         * config/aarch64/aarch64-simd-builtins.def
30150         (aarch64_sm3ss1qv4si): Ditto.
30151         (aarch64_sm3tt1aq4si): Ditto.
30152         (aarch64_sm3tt1bq4si): Ditto.
30153         (aarch64_sm3tt2aq4si): Ditto.
30154         (aarch64_sm3tt2bq4si): Ditto.
30155         (aarch64_sm3partw1qv4si): Ditto.
30156         (aarch64_sm3partw2qv4si): Ditto.
30157         (aarch64_sm4eqv4si): Ditto.
30158         (aarch64_sm4ekeyqv4si): Ditto.
30159         * config/aarch64/aarch64-simd.md:
30160         (aarch64_sm3ss1qv4si): Ditto.
30161         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
30162         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
30163         (aarch64_sm4eqv4si): Ditto.
30164         (aarch64_sm4ekeyqv4si): Ditto.
30165         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
30166         (sm3part_op): Ditto.
30167         (CRYPTO_SM3TT): Ditto.
30168         (CRYPTO_SM3PART): Ditto.
30169         (UNSPEC_SM3SS1): New unspec.
30170         (UNSPEC_SM3TT1A): Ditto.
30171         (UNSPEC_SM3TT1B): Ditto.
30172         (UNSPEC_SM3TT2A): Ditto.
30173         (UNSPEC_SM3TT2B): Ditto.
30174         (UNSPEC_SM3PARTW1): Ditto.
30175         (UNSPEC_SM3PARTW2): Ditto.
30176         (UNSPEC_SM4E): Ditto.
30177         (UNSPEC_SM4EKEY): Ditto.
30178         * config/aarch64/constraints.md (Ui2): New constraint.
30179         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
30180         * config/arm/types.md (crypto_sm3): New type attribute.
30181         (crypto_sm4): Ditto.
30182         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
30183         (vsm3tt1aq_u32): Ditto.
30184         (vsm3tt1bq_u32): Ditto.
30185         (vsm3tt2aq_u32): Ditto.
30186         (vsm3tt2bq_u32): Ditto.
30187         (vsm3partw1q_u32): Ditto.
30188         (vsm3partw2q_u32): Ditto.
30189         (vsm4eq_u32): Ditto.
30190         (vsm4ekeyq_u32): Ditto.
30191         (doc/invoke.texi): Document new sm4 option.
30193 2018-01-10  Michael Collison  <michael.collison@arm.com>
30195         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
30196         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
30197         (AARCH64_FL_FOR_ARCH8_4): New.
30198         (AARCH64_FL_V8_4): New flag.
30199         (doc/invoke.texi): Document new armv8.4-a option.
30201 2018-01-10  Michael Collison  <michael.collison@arm.com>
30203         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30204         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
30205         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
30206         * config/aarch64/aarch64-option-extension.def: Add
30207         AARCH64_OPT_EXTENSION of 'sha2'.
30208         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
30209         (crypto): Disable sha2 and aes if crypto disabled.
30210         (crypto): Enable aes and sha2 if enabled.
30211         (simd): Disable sha2 and aes if simd disabled.
30212         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
30213         New flags.
30214         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
30215         (TARGET_SHA2): New feature flag for sha2.
30216         (TARGET_AES): New feature flag for aes.
30217         * config/aarch64/aarch64-simd.md:
30218         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
30219         conditional on TARGET_AES.
30220         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
30221         (aarch64_crypto_sha1hsi): Make pattern conditional
30222         on TARGET_SHA2.
30223         (aarch64_crypto_sha1hv4si): Ditto.
30224         (aarch64_be_crypto_sha1hv4si): Ditto.
30225         (aarch64_crypto_sha1su1v4si): Ditto.
30226         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
30227         (aarch64_crypto_sha1su0v4si): Ditto.
30228         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
30229         (aarch64_crypto_sha256su0v4si): Ditto.
30230         (aarch64_crypto_sha256su1v4si): Ditto.
30231         (doc/invoke.texi): Document new aes and sha2 options.
30233 2018-01-10  Martin Sebor  <msebor@redhat.com>
30235         PR tree-optimization/83781
30236         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
30237         as string arrays.
30239 2018-01-11  Martin Sebor  <msebor@gmail.com>
30240             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
30242         PR tree-optimization/83501
30243         PR tree-optimization/81703
30245         * tree-ssa-strlen.c (get_string_cst): Rename...
30246         (get_string_len): ...to this.  Handle global constants.
30247         (handle_char_store): Adjust.
30249 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
30250             Jim Wilson  <jimw@sifive.com>
30252         * config/riscv/riscv-protos.h (riscv_output_return): New.
30253         * config/riscv/riscv.c (struct machine_function): New naked_p field.
30254         (riscv_attribute_table, riscv_output_return),
30255         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
30256         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
30257         (riscv_compute_frame_info): Only compute frame->mask if not a naked
30258         function.
30259         (riscv_expand_prologue): Add early return for naked function.
30260         (riscv_expand_epilogue): Likewise.
30261         (riscv_function_ok_for_sibcall): Return false for naked function.
30262         (riscv_set_current_function): New.
30263         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
30264         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
30265         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
30266         * doc/extend.texi (RISC-V Function Attributes): New.
30268 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
30270         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
30271         check for 128-bit long double before checking TCmode.
30272         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
30273         128-bit long doubles before checking TFmode or TCmode.
30274         (FLOAT128_IBM_P): Likewise.
30276 2018-01-10  Martin Sebor  <msebor@redhat.com>
30278         PR tree-optimization/83671
30279         * builtins.c (c_strlen): Unconditionally return zero for the empty
30280         string.
30281         Use -Warray-bounds for warnings.
30282         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
30283         for non-constant array indices with COMPONENT_REF, arrays of
30284         arrays, and pointers to arrays.
30285         (gimple_fold_builtin_strlen): Determine and set length range for
30286         non-constant character arrays.
30288 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
30290         PR middle-end/81897
30291         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
30292         empty blocks.
30294 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
30296         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
30298 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
30300         PR target/83399
30301         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
30302         VECTOR_MEM_ALTIVEC_OR_VSX_P.
30303         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
30304         indexed_or_indirect_operand predicate.
30305         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
30306         (*vsx_le_perm_load_v8hi): Likewise.
30307         (*vsx_le_perm_load_v16qi): Likewise.
30308         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
30309         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
30310         (*vsx_le_perm_store_v8hi): Likewise.
30311         (*vsx_le_perm_store_v16qi): Likewise.
30312         (eight unnamed splitters): Likewise.
30314 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
30316         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
30317         * config/rs6000/emmintrin.h: Likewise.
30318         * config/rs6000/mmintrin.h: Likewise.
30319         * config/rs6000/xmmintrin.h: Likewise.
30321 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
30323         PR c++/43486
30324         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
30325         "public_flag".
30326         * tree.c (tree_nop_conversion): Return true for location wrapper
30327         nodes.
30328         (maybe_wrap_with_location): New function.
30329         (selftest::check_strip_nops): New function.
30330         (selftest::test_location_wrappers): New function.
30331         (selftest::tree_c_tests): Call it.
30332         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
30333         (maybe_wrap_with_location): New decl.
30334         (EXPR_LOCATION_WRAPPER_P): New macro.
30335         (location_wrapper_p): New inline function.
30336         (tree_strip_any_location_wrapper): New inline function.
30338 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
30340         PR target/83735
30341         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
30342         stack_realign_offset for the largest alignment of stack slot
30343         actually used.
30344         (ix86_find_max_used_stack_alignment): New function.
30345         (ix86_finalize_stack_frame_flags): Use it.  Set
30346         max_used_stack_alignment if we don't realign stack.
30347         * config/i386/i386.h (machine_function): Add
30348         max_used_stack_alignment.
30350 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
30352         * config/arm/arm.opt (-mbranch-cost): New option.
30353         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
30354         account.
30356 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
30358         PR target/83629
30359         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
30360         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
30362 2018-01-10  Richard Biener  <rguenther@suse.de>
30364         PR debug/83765
30365         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
30366         early out so it also covers the case where we have a non-NULL
30367         origin.
30369 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
30371         PR tree-optimization/83753
30372         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
30373         for non-strided grouped accesses if the number of elements is 1.
30375 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
30377         PR target/81616
30378         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
30379         * i386.h (TARGET_USE_GATHER): Define.
30380         * x86-tune.def (X86_TUNE_USE_GATHER): New.
30382 2018-01-10  Martin Liska  <mliska@suse.cz>
30384         PR bootstrap/82831
30385         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
30386         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
30387         partitioning.
30388         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
30389         CLEANUP_NO_PARTITIONING is not set.
30391 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
30393         * doc/rtl.texi: Remove documentation of (const ...) wrappers
30394         for vectors, as a partial revert of r254296.
30395         * rtl.h (const_vec_p): Delete.
30396         (const_vec_duplicate_p): Don't test for vector CONSTs.
30397         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
30398         * expmed.c (make_tree): Likewise.
30400         Revert:
30401         * common.md (E, F): Use CONSTANT_P instead of checking for
30402         CONST_VECTOR.
30403         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
30404         checking for CONST_VECTOR.
30406 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
30408         PR middle-end/83575
30409         * predict.c (force_edge_cold): Handle in more sane way edges
30410         with no prediction.
30412 2018-01-09  Carl Love  <cel@us.ibm.com>
30414         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
30415         V4SI, V4SF types.
30416         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
30417         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
30418         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
30419         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
30420         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
30421         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
30422         * config/rs6000/rs6000-protos.h: Add extern defition for
30423         rs6000_generate_float2_double_code.
30424         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
30425         function.
30426         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
30427         (float2_v2df): Add define_expand.
30429 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
30431         PR target/83628
30432         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
30433         op_mode in the force_to_mode call.
30435 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
30437         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
30438         instead of checking each element individually.
30439         (aarch64_evpc_uzp): Likewise.
30440         (aarch64_evpc_zip): Likewise.
30441         (aarch64_evpc_ext): Likewise.
30442         (aarch64_evpc_rev): Likewise.
30443         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
30444         instead of checking each element individually.  Return true without
30445         generating rtl if
30446         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
30447         whether all selected elements come from the same input, instead of
30448         checking each element individually.  Remove calls to gen_rtx_REG,
30449         start_sequence and end_sequence and instead assert that no rtl is
30450         generated.
30452 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
30454         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
30455         order of HIGH and CONST checks.
30457 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
30459         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
30460         if the destination isn't an SSA_NAME.
30462 2018-01-09  Richard Biener  <rguenther@suse.de>
30464         PR tree-optimization/83668
30465         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
30466         move prologue...
30467         (canonicalize_loop_form): ... here, renamed from ...
30468         (canonicalize_loop_closed_ssa_form): ... this and amended to
30469         swap successor edges for loop exit blocks to make us use
30470         the RPO order we need for initial schedule generation.
30472 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
30474         PR tree-optimization/64811
30475         * match.pd: When optimizing comparisons with Inf, avoid
30476         introducing or losing exceptions from comparisons with NaN.
30478 2018-01-09  Martin Liska  <mliska@suse.cz>
30480         PR sanitizer/82517
30481         * asan.c (shadow_mem_size): Add gcc_assert.
30483 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
30485         Don't save registers in main().
30487         PR target/83738
30488         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
30489         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
30490         * config/avr/avr.c (avr_set_current_function): Don't error if
30491         naked, OS_task or OS_main are specified at the same time.
30492         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
30493         OS_main.
30494         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
30495         attribute.
30496         * common/config/avr/avr-common.c (avr_option_optimization_table):
30497         Switch on -mmain-is-OS_task for optimizing compilations.
30499 2018-01-09  Richard Biener  <rguenther@suse.de>
30501         PR tree-optimization/83572
30502         * graphite.c: Include cfganal.h.
30503         (graphite_transform_loops): Connect infinite loops to exit
30504         and remove fake edges at the end.
30506 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
30508         * ipa-inline.c (edge_badness): Revert accidental checkin.
30510 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
30512         PR ipa/80763
30513         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
30514         symbols; not inline clones.
30516 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
30518         PR target/83507
30519         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
30520         hard registers.  Formatting fixes.
30522         PR preprocessor/83722
30523         * gcc.c (try_generate_repro): Pass
30524         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
30525         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
30526         do_report_bug.
30528 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
30529             Kito Cheng  <kito.cheng@gmail.com>
30531         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
30532         (riscv_leaf_function_p): Delete.
30533         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
30535 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
30537         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
30538         function.
30539         (do_ifelse): New function.
30540         (do_isel): New function.
30541         (do_sub3): New function.
30542         (do_add3): New function.
30543         (do_load_mask_compare): New function.
30544         (do_overlap_load_compare): New function.
30545         (expand_compare_loop): New function.
30546         (expand_block_compare): Call expand_compare_loop() when appropriate.
30547         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
30548         option description.
30549         (-mblock-compare-inline-loop-limit): New option.
30551 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
30553         PR target/83677
30554         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
30555         Reverse order of second and third operands in first alternative.
30556         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
30557         of first and second elements in UNSPEC_VPERMR vector.
30558         (altivec_expand_vec_perm_le): Likewise.
30560 2018-01-08  Jeff Law  <law@redhat.com>
30562         PR rtl-optimizatin/81308
30563         * tree-switch-conversion.c (cfg_altered): New file scoped static.
30564         (process_switch): If group_case_labels makes a change, then set
30565         cfg_altered.
30566         (pass_convert_switch::execute): If a switch is converted, then
30567         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
30569         PR rtl-optimization/81308
30570         * recog.c (split_all_insns): Conditionally cleanup the CFG after
30571         splitting insns.
30573 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
30575         PR target/83663 - Revert r255946
30576         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
30577         generation for cases where splatting a value is not useful.
30578         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
30579         across a vec_duplicate and a paradoxical subreg forming a vector
30580         mode to a vec_concat.
30582 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
30584         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
30585         -march=armv8.3-a variants.
30586         * config/arm/t-multilib: Likewise.
30587         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
30589 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
30591         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
30592         to generate rtl.
30593         (cceq_ior_compare_complement): Give it a name so I can use it, and
30594         change boolean_or_operator predicate to boolean_operator so it can
30595         be used to generate a crand.
30596         (eqne): New code iterator.
30597         (bd/bd_neg): New code_attrs.
30598         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
30599         a single define_insn.
30600         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
30601         decrement (bdnzt/bdnzf/bdzt/bdzf).
30602         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
30603         with the new names of the branch decrement patterns, and added the
30604         names of the branch decrement conditional patterns.
30606 2018-01-08  Richard Biener  <rguenther@suse.de>
30608         PR tree-optimization/83563
30609         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
30610         cache.
30612 2018-01-08  Richard Biener  <rguenther@suse.de>
30614         PR middle-end/83713
30615         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
30617 2018-01-08  Richard Biener  <rguenther@suse.de>
30619         PR tree-optimization/83685
30620         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
30621         references to abnormals.
30623 2018-01-08  Richard Biener  <rguenther@suse.de>
30625         PR lto/83719
30626         * dwarf2out.c (output_indirect_strings): Handle empty
30627         skeleton_debug_str_hash.
30628         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
30630 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
30632         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
30633         (emit_store_direct): Likewise.
30634         (arc_trampoline_adjust_address): Likewise.
30635         (arc_asm_trampoline_template): New function.
30636         (arc_initialize_trampoline): Use asm_trampoline_template.
30637         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
30638         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
30639         * config/arc/arc.md (flush_icache): Delete pattern.
30641 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
30643         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
30644         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
30645         munaligned-access.
30647 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30649         PR target/83681
30650         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
30651         by not USED_FOR_TARGET.
30652         (make_pass_resolve_sw_modes): Likewise.
30654 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30656         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
30657         USED_FOR_TARGET.
30659 2018-01-08  Richard Biener  <rguenther@suse.de>
30661         PR middle-end/83580
30662         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
30664 2018-01-08  Richard Biener  <rguenther@suse.de>
30666         PR middle-end/83517
30667         * match.pd ((t * 2) / 2) -> t): Add missing :c.
30669 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
30671         PR middle-end/81897
30672         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
30673         basic blocks with a small number of successors.
30674         (convert_control_dep_chain_into_preds): Improve handling of
30675         forwarder blocks.
30676         (dump_predicates): Split apart into...
30677         (dump_pred_chain): ...here...
30678         (dump_pred_info): ...and here.
30679         (can_one_predicate_be_invalidated_p): Add debugging printfs.
30680         (can_chain_union_be_invalidated_p): Improve check for invalidation
30681         of paths.
30682         (uninit_uses_cannot_happen): Avoid unnecessary if
30683         convert_control_dep_chain_into_preds yielded nothing.
30685 2018-01-06  Martin Sebor  <msebor@redhat.com>
30687         PR tree-optimization/83640
30688         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
30689         subtracting negative offset from size.
30690         (builtin_access::overlap): Adjust offset bounds of the access to fall
30691         within the size of the object if possible.
30693 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
30695         PR rtl-optimization/83699
30696         * expmed.c (extract_bit_field_1): Restrict the vector usage of
30697         extract_bit_field_as_subreg to cases in which the extracted
30698         value is also a vector.
30700         * lra-constraints.c (process_alt_operands): Test for the equivalence
30701         substitutions when detecting a possible reload cycle.
30703 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
30705         PR debug/83480
30706         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
30707         by default if flag_selective_schedling{,2}.  Formatting fixes.
30709         PR rtl-optimization/83682
30710         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
30711         if it has non-VECTOR_MODE element mode.
30712         (vec_duplicate_p): Likewise.
30714         PR middle-end/83694
30715         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
30716         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
30718 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
30720         PR target/83604
30721         * config/i386/i386-builtin.def
30722         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
30723         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
30724         Require also OPTION_MASK_ISA_AVX512F in addition to
30725         OPTION_MASK_ISA_GFNI.
30726         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
30727         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
30728         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
30729         to OPTION_MASK_ISA_GFNI.
30730         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
30731         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
30732         OPTION_MASK_ISA_AVX512BW.
30733         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
30734         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
30735         addition to OPTION_MASK_ISA_GFNI.
30736         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
30737         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
30738         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
30739         to OPTION_MASK_ISA_GFNI.
30740         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
30741         a requirement for all ISAs rather than any of them with a few
30742         exceptions.
30743         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
30744         processing.
30745         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
30746         bitmasks to be enabled with 3 exceptions, instead of requiring any
30747         enabled ISA with lots of exceptions.
30748         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
30749         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
30750         Change avx512bw in isa attribute to avx512f.
30751         * config/i386/sgxintrin.h: Add license boilerplate.
30752         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
30753         to __AVX512F__ and __AVX512VL to __AVX512VL__.
30754         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
30755         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
30756         defined.
30757         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
30758         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
30759         temporarily sse2 rather than sse if not enabled already.
30761         PR target/83604
30762         * config/i386/sse.md (VI248_VLBW): Rename to ...
30763         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
30764         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
30765         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
30766         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
30767         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
30768         mode iterator instead of VI248_VLBW.
30770 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
30772         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
30773         (record_modified): Skip clobbers; add debug output.
30774         (param_change_prob): Use sreal frequencies.
30776 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
30778         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
30779         punt for user-aligned variables.
30781 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
30783         * tree-chrec.c (chrec_contains_symbols): Return true for
30784         POLY_INT_CST.
30786 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
30788         PR target/82439
30789         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
30790         of (x|y) == x for BICS pattern.
30792 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
30794         PR tree-optimization/83605
30795         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
30796         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
30797         can throw.
30799 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30801         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
30802         * config/epiphany/rtems.h: New file.
30804 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30805             Uros Bizjak  <ubizjak@gmail.com>
30807         PR target/83554
30808         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
30809         QIreg_operand instead of register_operand predicate.
30810         * config/i386/i386.c (ix86_rop_should_change_byte_p,
30811         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
30812         comments instead of -fmitigate[-_]rop.
30814 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
30816         PR bootstrap/81926
30817         * cgraphunit.c (symbol_table::compile): Switch to text_section
30818         before calling assembly_start debug hook.
30819         * run-rtl-passes.c (run_rtl_passes): Likewise.
30820         Include output.h.
30822 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30824         * tree-vrp.c (extract_range_from_binary_expr_1): Check
30825         range_int_cst_p rather than !symbolic_range_p before calling
30826         extract_range_from_multiplicative_op_1.
30828 2018-01-04  Jeff Law  <law@redhat.com>
30830         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
30831         redundant test in assertion.
30833 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30835         * doc/rtl.texi: Document machine_mode wrapper classes.
30837 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30839         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
30840         using tree_to_uhwi.
30842 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30844         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
30845         the VEC_PERM_EXPR fold to fail.
30847 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30849         PR debug/83585
30850         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
30851         to switched_sections.
30853 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30855         PR target/83680
30856         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
30857         test for d.testing.
30859 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
30861         PR target/83387
30862         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
30863         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
30865 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30867         PR debug/83666
30868         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
30869         is BLKmode and bitpos not zero or mode change is needed.
30871 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30873         PR target/83675
30874         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
30875         TARGET_VIS2.
30877 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
30879         PR target/83628
30880         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
30881         instead of MULT rtx.  Update all corresponding splitters.
30882         (*saddl_se): Ditto.
30883         (*ssub<modesuffix>): Ditto.
30884         (*ssubl_se): Ditto.
30885         (*cmp_sadd_di): Update split patterns.
30886         (*cmp_sadd_si): Ditto.
30887         (*cmp_sadd_sidi): Ditto.
30888         (*cmp_ssub_di): Ditto.
30889         (*cmp_ssub_si): Ditto.
30890         (*cmp_ssub_sidi): Ditto.
30891         * config/alpha/predicates.md (const23_operand): New predicate.
30892         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
30893         Look for ASHIFT, not MULT inner operand.
30894         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
30896 2018-01-04  Martin Liska  <mliska@suse.cz>
30898         PR gcov-profile/83669
30899         * gcov.c (output_intermediate_file): Add version to intermediate
30900         gcov file.
30901         * doc/gcov.texi: Document new field 'version' in intermediate
30902         file format. Fix location of '-k' option of gcov command.
30904 2018-01-04  Martin Liska  <mliska@suse.cz>
30906         PR ipa/82352
30907         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
30909 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30911         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
30913 2018-01-03  Martin Sebor  <msebor@redhat.com>
30915         PR tree-optimization/83655
30916         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
30917         checking calls with invalid arguments.
30919 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30921         * tree-vect-stmts.c (vect_get_store_rhs): New function.
30922         (vectorizable_mask_load_store): Delete.
30923         (vectorizable_call): Return false for masked loads and stores.
30924         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
30925         instead of gimple_assign_rhs1.
30926         (vectorizable_load): Handle IFN_MASK_LOAD.
30927         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
30929 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30931         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
30932         split out from..,
30933         (vectorizable_mask_load_store): ...here.
30934         (vectorizable_load): ...and here.
30936 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30938         * tree-vect-stmts.c (vect_build_all_ones_mask)
30939         (vect_build_zero_merge_argument): New functions, split out from...
30940         (vectorizable_load): ...here.
30942 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30944         * tree-vect-stmts.c (vect_check_store_rhs): New function,
30945         split out from...
30946         (vectorizable_mask_load_store): ...here.
30947         (vectorizable_store): ...and here.
30949 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30951         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
30952         split out from...
30953         (vectorizable_mask_load_store): ...here.
30955 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30957         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
30958         (vect_model_store_cost): Take a vec_load_store_type instead of a
30959         vect_def_type.
30960         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
30961         (vect_model_store_cost): Take a vec_load_store_type instead of a
30962         vect_def_type.
30963         (vectorizable_mask_load_store): Update accordingly.
30964         (vectorizable_store): Likewise.
30965         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
30967 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30969         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
30970         IFN_MASK_LOAD calls here rather than...
30971         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
30973 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30974             Alan Hayward  <alan.hayward@arm.com>
30975             David Sherwood  <david.sherwood@arm.com>
30977         * expmed.c (extract_bit_field_1): For vector extracts,
30978         fall back to extract_bit_field_as_subreg if vec_extract
30979         isn't available.
30981 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30982             Alan Hayward  <alan.hayward@arm.com>
30983             David Sherwood  <david.sherwood@arm.com>
30985         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
30986         they are variable or constant sized.
30987         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
30988         slots for constant-sized data.
30990 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30991             Alan Hayward  <alan.hayward@arm.com>
30992             David Sherwood  <david.sherwood@arm.com>
30994         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
30995         handling COND_EXPRs with boolean comparisons, try to find a better
30996         basis for the mask type than the boolean itself.
30998 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31000         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
31001         is calculated and how it can be overridden.
31002         * genmodes.c (max_bitsize_mode_any_mode): New variable.
31003         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
31004         if defined.
31005         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
31006         if nonzero.
31008 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31009             Alan Hayward  <alan.hayward@arm.com>
31010             David Sherwood  <david.sherwood@arm.com>
31012         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
31013         Remove the mode argument.
31014         (aarch64_simd_valid_immediate): Remove the mode and inverse
31015         arguments.
31016         * config/aarch64/iterators.md (bitsize): New iterator.
31017         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
31018         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
31019         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
31020         aarch64_simd_valid_immediate.
31021         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
31022         (aarch64_reg_or_bic_imm): Likewise.
31023         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
31024         with an insn_type enum and msl with a modifier_type enum.
31025         Replace element_width with a scalar_mode.  Change the shift
31026         to unsigned int.  Add constructors for scalar_float_mode and
31027         scalar_int_mode elements.
31028         (aarch64_vect_float_const_representable_p): Delete.
31029         (aarch64_can_const_movi_rtx_p)
31030         (aarch64_simd_scalar_immediate_valid_for_move)
31031         (aarch64_simd_make_constant): Update call to
31032         aarch64_simd_valid_immediate.
31033         (aarch64_advsimd_valid_immediate_hs): New function.
31034         (aarch64_advsimd_valid_immediate): Likewise.
31035         (aarch64_simd_valid_immediate): Remove mode and inverse
31036         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
31037         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
31038         and aarch64_float_const_representable_p on the result.
31039         (aarch64_output_simd_mov_immediate): Remove mode argument.
31040         Update call to aarch64_simd_valid_immediate and use of
31041         simd_immediate_info.
31042         (aarch64_output_scalar_simd_mov_immediate): Update call
31043         accordingly.
31045 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31046             Alan Hayward  <alan.hayward@arm.com>
31047             David Sherwood  <david.sherwood@arm.com>
31049         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
31050         (mode_nunits): Likewise CONST_MODE_NUNITS.
31051         * machmode.def (ADJUST_NUNITS): Document.
31052         * genmodes.c (mode_data::need_nunits_adj): New field.
31053         (blank_mode): Update accordingly.
31054         (adj_nunits): New variable.
31055         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
31056         parameter.
31057         (emit_mode_size_inline): Set need_bytesize_adj for all modes
31058         listed in adj_nunits.
31059         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
31060         listed in adj_nunits.  Don't emit case statements for such modes.
31061         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
31062         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
31063         nothing if adj_nunits is nonnull.
31064         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
31065         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
31066         (emit_mode_fbit): Update use of print_maybe_const_decl.
31067         (emit_move_size): Likewise.  Treat the array as non-const
31068         if adj_nunits.
31069         (emit_mode_adjustments): Handle adj_nunits.
31071 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31073         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
31074         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
31075         (VECTOR_MODES): Use it.
31076         (make_vector_modes): Take the prefix as an argument.
31078 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31079             Alan Hayward  <alan.hayward@arm.com>
31080             David Sherwood  <david.sherwood@arm.com>
31082         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
31083         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
31084         for MODE_VECTOR_BOOL.
31085         * machmode.def (VECTOR_BOOL_MODE): Document.
31086         * genmodes.c (VECTOR_BOOL_MODE): New macro.
31087         (make_vector_bool_mode): New function.
31088         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
31089         MODE_VECTOR_BOOL.
31090         * lto-streamer-in.c (lto_input_mode_table): Likewise.
31091         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
31092         Likewise.
31093         * stor-layout.c (int_mode_for_mode): Likewise.
31094         * tree.c (build_vector_type_for_mode): Likewise.
31095         * varasm.c (output_constant_pool_2): Likewise.
31096         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
31097         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
31098         for MODE_VECTOR_BOOL.
31099         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
31100         of mode class checks.
31101         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
31102         instead of a list of mode class checks.
31103         (expand_vector_scalar_condition): Likewise.
31104         (type_for_widest_vector_mode): Handle BImode as an inner mode.
31106 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31107             Alan Hayward  <alan.hayward@arm.com>
31108             David Sherwood  <david.sherwood@arm.com>
31110         * machmode.h (mode_size): Change from unsigned short to
31111         poly_uint16_pod.
31112         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
31113         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
31114         or if measurement_type is not polynomial.
31115         (fixed_size_mode::includes_p): Check for constant-sized modes.
31116         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
31117         return a poly_uint16 rather than an unsigned short.
31118         (emit_mode_size): Change the type of mode_size from unsigned short
31119         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
31120         (emit_mode_adjustments): Cope with polynomial vector sizes.
31121         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
31122         for GET_MODE_SIZE.
31123         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
31124         for GET_MODE_SIZE.
31125         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
31126         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
31127         * caller-save.c (setup_save_areas): Likewise.
31128         (replace_reg_with_saved_mem): Likewise.
31129         * calls.c (emit_library_call_value_1): Likewise.
31130         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
31131         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
31132         (gen_lowpart_for_combine): Likewise.
31133         * convert.c (convert_to_integer_1): Likewise.
31134         * cse.c (equiv_constant, cse_insn): Likewise.
31135         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
31136         (cselib_subst_to_values): Likewise.
31137         * dce.c (word_dce_process_block): Likewise.
31138         * df-problems.c (df_word_lr_mark_ref): Likewise.
31139         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
31140         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
31141         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
31142         (rtl_for_decl_location): Likewise.
31143         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
31144         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
31145         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
31146         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
31147         (expand_expr_real_1): Likewise.
31148         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
31149         (pad_below): Likewise.
31150         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
31151         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
31152         * ira.c (get_subreg_tracking_sizes): Likewise.
31153         * ira-build.c (ira_create_allocno_objects): Likewise.
31154         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
31155         (ira_sort_regnos_for_alter_reg): Likewise.
31156         * ira-costs.c (record_operand_costs): Likewise.
31157         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
31158         (resolve_simple_move): Likewise.
31159         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
31160         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
31161         (lra_constraints): Likewise.
31162         (CONST_POOL_OK_P): Reject variable-sized modes.
31163         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
31164         (add_pseudo_to_slot, lra_spill): Likewise.
31165         * omp-low.c (omp_clause_aligned_alignment): Likewise.
31166         * optabs-query.c (get_best_extraction_insn): Likewise.
31167         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
31168         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
31169         (expand_mult_highpart, valid_multiword_target_p): Likewise.
31170         * recog.c (offsettable_address_addr_space_p): Likewise.
31171         * regcprop.c (maybe_mode_change): Likewise.
31172         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
31173         * regrename.c (build_def_use): Likewise.
31174         * regstat.c (dump_reg_info): Likewise.
31175         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
31176         (find_reloads, find_reloads_subreg_address): Likewise.
31177         * reload1.c (eliminate_regs_1): Likewise.
31178         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
31179         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
31180         (simplify_binary_operation_1, simplify_subreg): Likewise.
31181         * targhooks.c (default_function_arg_padding): Likewise.
31182         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
31183         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
31184         (verify_gimple_assign_ternary): Likewise.
31185         * tree-inline.c (estimate_move_cost): Likewise.
31186         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31187         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
31188         (get_address_cost_ainc): Likewise.
31189         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
31190         (vect_supportable_dr_alignment): Likewise.
31191         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
31192         (vectorizable_reduction): Likewise.
31193         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
31194         (vectorizable_operation, vectorizable_load): Likewise.
31195         * tree.c (build_same_sized_truth_vector_type): Likewise.
31196         * valtrack.c (cleanup_auto_inc_dec): Likewise.
31197         * var-tracking.c (emit_note_insn_var_location): Likewise.
31198         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
31199         (ADDR_VEC_ALIGN): Likewise.
31201 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31202             Alan Hayward  <alan.hayward@arm.com>
31203             David Sherwood  <david.sherwood@arm.com>
31205         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
31206         unsigned short.
31207         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
31208         or if measurement_type is polynomial.
31209         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
31210         * combine.c (make_extraction): Likewise.
31211         * dse.c (find_shift_sequence): Likewise.
31212         * dwarf2out.c (mem_loc_descriptor): Likewise.
31213         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
31214         (extract_bit_field, extract_low_bits): Likewise.
31215         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
31216         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
31217         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
31218         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
31219         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
31220         * reload.c (find_reloads): Likewise.
31221         * reload1.c (alter_reg): Likewise.
31222         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
31223         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
31224         * tree-if-conv.c (predicate_mem_writes): Likewise.
31225         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
31226         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
31227         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
31228         * valtrack.c (dead_debug_insert_temp): Likewise.
31229         * varasm.c (mergeable_constant_section): Likewise.
31230         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
31232 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31233             Alan Hayward  <alan.hayward@arm.com>
31234             David Sherwood  <david.sherwood@arm.com>
31236         * expr.c (expand_assignment): Cope with polynomial mode sizes
31237         when assigning to a CONCAT.
31239 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31240             Alan Hayward  <alan.hayward@arm.com>
31241             David Sherwood  <david.sherwood@arm.com>
31243         * machmode.h (mode_precision): Change from unsigned short to
31244         poly_uint16_pod.
31245         (mode_to_precision): Return a poly_uint16 rather than an unsigned
31246         short.
31247         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
31248         or if measurement_type is not polynomial.
31249         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
31250         in which the mode is already known to be a scalar_int_mode.
31251         * genmodes.c (emit_mode_precision): Change the type of mode_precision
31252         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
31253         initializer.
31254         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
31255         for GET_MODE_PRECISION.
31256         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
31257         for GET_MODE_PRECISION.
31258         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
31259         as polynomial.
31260         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
31261         (expand_field_assignment, make_extraction): Likewise.
31262         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
31263         (get_last_value): Likewise.
31264         * convert.c (convert_to_integer_1): Likewise.
31265         * cse.c (cse_insn): Likewise.
31266         * expr.c (expand_expr_real_1): Likewise.
31267         * lra-constraints.c (simplify_operand_subreg): Likewise.
31268         * optabs-query.c (can_atomic_load_p): Likewise.
31269         * optabs.c (expand_atomic_load): Likewise.
31270         (expand_atomic_store): Likewise.
31271         * ree.c (combine_reaching_defs): Likewise.
31272         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
31273         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
31274         * tree.h (type_has_mode_precision_p): Likewise.
31275         * ubsan.c (instrument_si_overflow): Likewise.
31277 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31278             Alan Hayward  <alan.hayward@arm.com>
31279             David Sherwood  <david.sherwood@arm.com>
31281         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
31282         polynomial numbers of units.
31283         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
31284         (valid_vector_subparts_p): New function.
31285         (build_vector_type): Remove temporary shim and take the number
31286         of units as a poly_uint64 rather than an int.
31287         (build_opaque_vector_type): Take the number of units as a
31288         poly_uint64 rather than an int.
31289         * tree.c (build_vector_from_ctor): Handle polynomial
31290         TYPE_VECTOR_SUBPARTS.
31291         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
31292         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
31293         (build_vector_from_val): If the number of units is variable,
31294         use build_vec_duplicate_cst for constant operands and
31295         VEC_DUPLICATE_EXPR otherwise.
31296         (make_vector_type): Remove temporary is_constant ().
31297         (build_vector_type, build_opaque_vector_type): Take the number of
31298         units as a poly_uint64 rather than an int.
31299         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
31300         VECTOR_CST_NELTS.
31301         * cfgexpand.c (expand_debug_expr): Likewise.
31302         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
31303         (store_constructor, expand_expr_real_1): Likewise.
31304         (const_scalar_mask_from_tree): Likewise.
31305         * fold-const-call.c (fold_const_reduction): Likewise.
31306         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
31307         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
31308         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
31309         (fold_relational_const): Likewise.
31310         (native_interpret_vector): Likewise.  Change the size from an
31311         int to an unsigned int.
31312         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
31313         TYPE_VECTOR_SUBPARTS.
31314         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
31315         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
31316         duplicating a non-constant operand into a variable-length vector.
31317         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
31318         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
31319         * ipa-icf.c (sem_variable::equals): Likewise.
31320         * match.pd: Likewise.
31321         * omp-simd-clone.c (simd_clone_subparts): Likewise.
31322         * print-tree.c (print_node): Likewise.
31323         * stor-layout.c (layout_type): Likewise.
31324         * targhooks.c (default_builtin_vectorization_cost): Likewise.
31325         * tree-cfg.c (verify_gimple_comparison): Likewise.
31326         (verify_gimple_assign_binary): Likewise.
31327         (verify_gimple_assign_ternary): Likewise.
31328         (verify_gimple_assign_single): Likewise.
31329         * tree-pretty-print.c (dump_generic_node): Likewise.
31330         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31331         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
31332         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
31333         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
31334         (vect_shift_permute_load_chain): Likewise.
31335         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
31336         (expand_vector_condition, optimize_vector_constructor): Likewise.
31337         (lower_vec_perm, get_compute_type): Likewise.
31338         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
31339         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
31340         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
31341         (vect_recog_mask_conversion_pattern): Likewise.
31342         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
31343         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
31344         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31345         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
31346         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
31347         (vectorizable_shift, vectorizable_operation, vectorizable_store)
31348         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
31349         (supportable_widening_operation): Likewise.
31350         (supportable_narrowing_operation): Likewise.
31351         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
31352         Likewise.
31353         * varasm.c (output_constant): Likewise.
31355 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31356             Alan Hayward  <alan.hayward@arm.com>
31357             David Sherwood  <david.sherwood@arm.com>
31359         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
31360         so that both the length == 3 and length != 3 cases set up their
31361         own permute vectors.  Add comments explaining why we know the
31362         number of elements is constant.
31363         (vect_permute_load_chain): Likewise.
31365 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31366             Alan Hayward  <alan.hayward@arm.com>
31367             David Sherwood  <david.sherwood@arm.com>
31369         * machmode.h (mode_nunits): Change from unsigned char to
31370         poly_uint16_pod.
31371         (ONLY_FIXED_SIZE_MODES): New macro.
31372         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
31373         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
31374         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
31375         New typedefs.
31376         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
31377         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
31378         or if measurement_type is not polynomial.
31379         * genmodes.c (ZERO_COEFFS): New macro.
31380         (emit_mode_nunits_inline): Make mode_nunits_inline return a
31381         poly_uint16.
31382         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
31383         Use ZERO_COEFFS when emitting initializers.
31384         * data-streamer.h (bp_pack_poly_value): New function.
31385         (bp_unpack_poly_value): Likewise.
31386         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
31387         for GET_MODE_NUNITS.
31388         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
31389         for GET_MODE_NUNITS.
31390         * tree.c (make_vector_type): Remove temporary shim and make
31391         the real function take the number of units as a poly_uint64
31392         rather than an int.
31393         (build_vector_type_for_mode): Handle polynomial nunits.
31394         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
31395         * emit-rtl.c (const_vec_series_p_1): Likewise.
31396         (gen_rtx_CONST_VECTOR): Likewise.
31397         * fold-const.c (test_vec_duplicate_folding): Likewise.
31398         * genrecog.c (validate_pattern): Likewise.
31399         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
31400         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
31401         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
31402         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
31403         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
31404         * rtlanal.c (subreg_get_info): Likewise.
31405         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
31406         (vect_grouped_load_supported): Likewise.
31407         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
31408         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
31409         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
31410         (simplify_const_unary_operation, simplify_binary_operation_1)
31411         (simplify_const_binary_operation, simplify_ternary_operation)
31412         (test_vector_ops_duplicate, test_vector_ops): Likewise.
31413         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
31414         instead of CONST_VECTOR_NUNITS.
31415         * varasm.c (output_constant_pool_2): Likewise.
31416         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
31417         explicit-encoded elements in the XVEC for variable-length vectors.
31419 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31421         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
31423 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31424             Alan Hayward  <alan.hayward@arm.com>
31425             David Sherwood  <david.sherwood@arm.com>
31427         * coretypes.h (fixed_size_mode): Declare.
31428         (fixed_size_mode_pod): New typedef.
31429         * builtins.h (target_builtins::x_apply_args_mode)
31430         (target_builtins::x_apply_result_mode): Change type to
31431         fixed_size_mode_pod.
31432         * builtins.c (apply_args_size, apply_result_size, result_vector)
31433         (expand_builtin_apply_args_1, expand_builtin_apply)
31434         (expand_builtin_return): Update accordingly.
31436 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31438         * cse.c (hash_rtx_cb): Hash only the encoded elements.
31439         * cselib.c (cselib_hash_rtx): Likewise.
31440         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
31441         CONST_VECTOR encoding.
31443 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31444             Jeff Law  <law@redhat.com>
31446         PR target/83641
31447         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
31448         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
31449         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
31450         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
31452         PR target/83641
31453         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
31454         explicitly probe *sp in a noreturn function if there were any callee
31455         register saves or frame pointer is needed.
31457 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31459         PR debug/83621
31460         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
31461         BLKmode for ternary, binary or unary expressions.
31463         PR debug/83645
31464         * var-tracking.c (delete_vta_debug_insn): New inline function.
31465         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
31466         insns from get_insns () to NULL instead of each bb separately.
31467         Use delete_vta_debug_insn.  No longer static.
31468         (vt_debug_insns_local, variable_tracking_main_1): Adjust
31469         delete_vta_debug_insns callers.
31470         * rtl.h (delete_vta_debug_insns): Declare.
31471         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
31472         instead of variable_tracking_main.
31474 2018-01-03  Martin Sebor  <msebor@redhat.com>
31476         PR tree-optimization/83603
31477         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
31478         arguments past the endof the argument list in functions declared
31479         without a prototype.
31480         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
31481         Avoid checking when arguments are null.
31483 2018-01-03  Martin Sebor  <msebor@redhat.com>
31485         PR c/83559
31486         * doc/extend.texi (attribute const): Fix a typo.
31487         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
31488         issuing -Wsuggest-attribute for void functions.
31490 2018-01-03  Martin Sebor  <msebor@redhat.com>
31492         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
31493         offset_int::from instead of wide_int::to_shwi.
31494         (maybe_diag_overlap): Remove assertion.
31495         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
31496         * gimple-ssa-sprintf.c (format_directive): Same.
31497         (parse_directive): Same.
31498         (sprintf_dom_walker::compute_format_length): Same.
31499         (try_substitute_return_value): Same.
31501 2018-01-03  Jeff Law  <law@redhat.com>
31503         PR middle-end/83654
31504         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
31505         non-constant residual for zero at runtime and avoid probing in
31506         that case.  Reorganize code for trailing problem to mirror handling
31507         of the residual.
31509 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
31511         PR tree-optimization/83501
31512         * tree-ssa-strlen.c (get_string_cst): New.
31513         (handle_char_store): Call get_string_cst.
31515 2018-01-03  Martin Liska  <mliska@suse.cz>
31517         PR tree-optimization/83593
31518         * tree-ssa-strlen.c: Include tree-cfg.h.
31519         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
31520         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
31521         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
31522         to false.
31523         (strlen_dom_walker::before_dom_children): Call
31524         gimple_purge_dead_eh_edges. Dump tranformation with details
31525         dump flags.
31526         (strlen_dom_walker::before_dom_children): Update call by adding
31527         new argument cleanup_eh.
31528         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
31530 2018-01-03  Martin Liska  <mliska@suse.cz>
31532         PR ipa/83549
31533         * cif-code.def (VARIADIC_THUNK): New enum value.
31534         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
31535         thunks.
31537 2018-01-03  Jan Beulich  <jbeulich@suse.com>
31539         * sse.md (mov<mode>_internal): Tighten condition for when to use
31540         vmovdqu<ssescalarsize> for TI and OI modes.
31542 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31544         Update copyright years.
31546 2018-01-03  Martin Liska  <mliska@suse.cz>
31548         PR ipa/83594
31549         * ipa-visibility.c (function_and_variable_visibility): Skip
31550         functions with noipa attribure.
31552 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31554         * gcc.c (process_command): Update copyright notice dates.
31555         * gcov-dump.c (print_version): Ditto.
31556         * gcov.c (print_version): Ditto.
31557         * gcov-tool.c (print_version): Ditto.
31558         * gengtype.c (create_file): Ditto.
31559         * doc/cpp.texi: Bump @copying's copyright year.
31560         * doc/cppinternals.texi: Ditto.
31561         * doc/gcc.texi: Ditto.
31562         * doc/gccint.texi: Ditto.
31563         * doc/gcov.texi: Ditto.
31564         * doc/install.texi: Ditto.
31565         * doc/invoke.texi: Ditto.
31567 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31569         * vector-builder.h (vector_builder::m_full_nelts): Change from
31570         unsigned int to poly_uint64.
31571         (vector_builder::full_nelts): Update prototype accordingly.
31572         (vector_builder::new_vector): Likewise.
31573         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
31574         (vector_builder::operator ==): Likewise.
31575         (vector_builder::finalize): Likewise.
31576         * int-vector-builder.h (int_vector_builder::int_vector_builder):
31577         Take the number of elements as a poly_uint64 rather than an
31578         unsigned int.
31579         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
31580         from unsigned int to poly_uint64.
31581         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
31582         (vec_perm_indices::new_vector): Likewise.
31583         (vec_perm_indices::length): Likewise.
31584         (vec_perm_indices::nelts_per_input): Likewise.
31585         (vec_perm_indices::input_nelts): Likewise.
31586         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
31587         number of elements per input as a poly_uint64 rather than an
31588         unsigned int.  Use the original encoding for variable-length
31589         vectors, rather than clamping each individual element.
31590         For the second and subsequent elements in each pattern,
31591         clamp the step and base before clamping their sum.
31592         (vec_perm_indices::series_p): Handle polynomial element counts.
31593         (vec_perm_indices::all_in_range_p): Likewise.
31594         (vec_perm_indices_to_tree): Likewise.
31595         (vec_perm_indices_to_rtx): Likewise.
31596         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
31597         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
31598         (tree_vector_builder::new_binary_operation): Handle polynomial
31599         element counts.  Return false if we need to know the number
31600         of elements at compile time.
31601         * fold-const.c (fold_vec_perm): Punt if the number of elements
31602         isn't known at compile time.
31604 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31606         * vec-perm-indices.h (vec_perm_builder): Change element type
31607         from HOST_WIDE_INT to poly_int64.
31608         (vec_perm_indices::element_type): Update accordingly.
31609         (vec_perm_indices::clamp): Handle polynomial element_types.
31610         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
31611         (vec_perm_indices::all_in_range_p): Likewise.
31612         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
31613         than shwi trees.
31614         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
31615         polynomial vec_perm_indices element types.
31616         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
31617         * fold-const.c (fold_vec_perm): Likewise.
31618         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
31619         * tree-vect-generic.c (lower_vec_perm): Likewise.
31620         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
31621         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
31622         element type to HOST_WIDE_INT.
31624 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31625             Alan Hayward  <alan.hayward@arm.com>
31626             David Sherwood  <david.sherwood@arm.com>
31628         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
31629         rather than an int.  Use plus_constant.
31630         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
31631         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
31633 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31634             Alan Hayward  <alan.hayward@arm.com>
31635             David Sherwood  <david.sherwood@arm.com>
31637         * calls.c (emit_call_1, expand_call): Change struct_value_size from
31638         a HOST_WIDE_INT to a poly_int64.
31640 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31641             Alan Hayward  <alan.hayward@arm.com>
31642             David Sherwood  <david.sherwood@arm.com>
31644         * calls.c (load_register_parameters): Cope with polynomial
31645         mode sizes.  Require a constant size for BLKmode parameters
31646         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
31647         forces a parameter to be padded at the lsb end in order to
31648         fill a complete number of words, require the parameter size
31649         to be ordered wrt UNITS_PER_WORD.
31651 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31652             Alan Hayward  <alan.hayward@arm.com>
31653             David Sherwood  <david.sherwood@arm.com>
31655         * reload1.c (spill_stack_slot_width): Change element type
31656         from unsigned int to poly_uint64_pod.
31657         (alter_reg): Treat mode sizes as polynomial.
31659 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31660             Alan Hayward  <alan.hayward@arm.com>
31661             David Sherwood  <david.sherwood@arm.com>
31663         * reload.c (complex_word_subreg_p): New function.
31664         (reload_inner_reg_of_subreg, push_reload): Use it.
31666 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31667             Alan Hayward  <alan.hayward@arm.com>
31668             David Sherwood  <david.sherwood@arm.com>
31670         * lra-constraints.c (process_alt_operands): Reject matched
31671         operands whose sizes aren't ordered.
31672         (match_reload): Refer to this check here.
31674 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31675             Alan Hayward  <alan.hayward@arm.com>
31676             David Sherwood  <david.sherwood@arm.com>
31678         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
31679         that the mode size is in the set {1, 2, 4, 8, 16}.
31681 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31682             Alan Hayward  <alan.hayward@arm.com>
31683             David Sherwood  <david.sherwood@arm.com>
31685         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
31686         Use plus_constant instead of gen_rtx_PLUS.
31688 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31689             Alan Hayward  <alan.hayward@arm.com>
31690             David Sherwood  <david.sherwood@arm.com>
31692         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
31693         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
31694         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
31695         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
31696         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
31697         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
31698         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
31699         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
31700         * config/i386/i386.c (ix86_push_rounding): ...this new function.
31701         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
31702         a poly_int64.
31703         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
31704         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
31705         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
31706         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
31707         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
31708         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
31709         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
31710         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
31711         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
31712         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
31713         function.
31714         * expr.c (emit_move_resolve_push): Treat the input and result
31715         of PUSH_ROUNDING as a poly_int64.
31716         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
31717         (emit_push_insn): Likewise.
31718         * lra-eliminations.c (mark_not_eliminable): Likewise.
31719         * recog.c (push_operand): Likewise.
31720         * reload1.c (elimination_effects): Likewise.
31721         * rtlanal.c (nonzero_bits1): Likewise.
31722         * calls.c (store_one_arg): Likewise.  Require the padding to be
31723         known at compile time.
31725 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31726             Alan Hayward  <alan.hayward@arm.com>
31727             David Sherwood  <david.sherwood@arm.com>
31729         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
31730         Use plus_constant instead of gen_rtx_PLUS.
31732 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31733             Alan Hayward  <alan.hayward@arm.com>
31734             David Sherwood  <david.sherwood@arm.com>
31736         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
31737         rather than an int.
31739 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31740             Alan Hayward  <alan.hayward@arm.com>
31741             David Sherwood  <david.sherwood@arm.com>
31743         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
31744         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
31745         via stack temporaries.  Treat the mode size as polynomial too.
31747 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31748             Alan Hayward  <alan.hayward@arm.com>
31749             David Sherwood  <david.sherwood@arm.com>
31751         * expr.c (expand_expr_real_2): When handling conversions involving
31752         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
31753         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
31754         as a poly_uint64 too.
31756 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31757             Alan Hayward  <alan.hayward@arm.com>
31758             David Sherwood  <david.sherwood@arm.com>
31760         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
31762 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31763             Alan Hayward  <alan.hayward@arm.com>
31764             David Sherwood  <david.sherwood@arm.com>
31766         * combine.c (can_change_dest_mode): Handle polynomial
31767         REGMODE_NATURAL_SIZE.
31768         * expmed.c (store_bit_field_1): Likewise.
31769         * expr.c (store_constructor): Likewise.
31770         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
31771         and polynomial REGMODE_NATURAL_SIZE.
31772         (gen_lowpart_common): Likewise.
31773         * reginfo.c (record_subregs_of_mode): Likewise.
31774         * rtlanal.c (read_modify_subreg_p): Likewise.
31776 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31777             Alan Hayward  <alan.hayward@arm.com>
31778             David Sherwood  <david.sherwood@arm.com>
31780         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
31781         numbers of elements.
31783 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31784             Alan Hayward  <alan.hayward@arm.com>
31785             David Sherwood  <david.sherwood@arm.com>
31787         * match.pd: Cope with polynomial numbers of vector elements.
31789 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31790             Alan Hayward  <alan.hayward@arm.com>
31791             David Sherwood  <david.sherwood@arm.com>
31793         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
31794         in a POINTER_PLUS_EXPR.
31796 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31797             Alan Hayward  <alan.hayward@arm.com>
31798             David Sherwood  <david.sherwood@arm.com>
31800         * omp-simd-clone.c (simd_clone_subparts): New function.
31801         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
31802         (ipa_simd_modify_function_body): Likewise.
31804 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31805             Alan Hayward  <alan.hayward@arm.com>
31806             David Sherwood  <david.sherwood@arm.com>
31808         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
31809         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
31810         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
31811         (expand_vector_condition, vector_element): Likewise.
31812         (subparts_gt): New function.
31813         (get_compute_type): Use subparts_gt.
31814         (count_type_subparts): Delete.
31815         (expand_vector_operations_1): Use subparts_gt instead of
31816         count_type_subparts.
31818 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31819             Alan Hayward  <alan.hayward@arm.com>
31820             David Sherwood  <david.sherwood@arm.com>
31822         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
31823         (vect_compile_time_alias): ...this new function.  Do the calculation
31824         on poly_ints rather than trees.
31825         (vect_prune_runtime_alias_test_list): Update call accordingly.
31827 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31828             Alan Hayward  <alan.hayward@arm.com>
31829             David Sherwood  <david.sherwood@arm.com>
31831         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
31832         numbers of units.
31833         (vect_schedule_slp_instance): Likewise.
31835 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31836             Alan Hayward  <alan.hayward@arm.com>
31837             David Sherwood  <david.sherwood@arm.com>
31839         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
31840         constant and extern definitions for variable-length vectors.
31841         (vect_get_constant_vectors): Note that the number of units
31842         is known to be constant.
31844 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31845             Alan Hayward  <alan.hayward@arm.com>
31846             David Sherwood  <david.sherwood@arm.com>
31848         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
31849         of units as polynomial.  Choose between WIDE and NARROW based
31850         on multiple_p.
31852 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31853             Alan Hayward  <alan.hayward@arm.com>
31854             David Sherwood  <david.sherwood@arm.com>
31856         * tree-vect-stmts.c (simd_clone_subparts): New function.
31857         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
31859 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31860             Alan Hayward  <alan.hayward@arm.com>
31861             David Sherwood  <david.sherwood@arm.com>
31863         * tree-vect-stmts.c (vectorizable_call): Treat the number of
31864         vectors as polynomial.  Use build_index_vector for
31865         IFN_GOMP_SIMD_LANE.
31867 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31868             Alan Hayward  <alan.hayward@arm.com>
31869             David Sherwood  <david.sherwood@arm.com>
31871         * tree-vect-stmts.c (get_load_store_type): Treat the number of
31872         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
31873         for variable-length vectors.
31874         (vectorizable_mask_load_store): Treat the number of units as
31875         polynomial, asserting that it is constant if the condition has
31876         already been enforced.
31877         (vectorizable_store, vectorizable_load): Likewise.
31879 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31880             Alan Hayward  <alan.hayward@arm.com>
31881             David Sherwood  <david.sherwood@arm.com>
31883         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
31884         of units as polynomial.  Punt if we can't tell at compile time
31885         which vector contains the final result.
31887 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31888             Alan Hayward  <alan.hayward@arm.com>
31889             David Sherwood  <david.sherwood@arm.com>
31891         * tree-vect-loop.c (vectorizable_induction): Treat the number
31892         of units as polynomial.  Punt on SLP inductions.  Use an integer
31893         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
31894         cast of such a series for variable-length floating-point
31895         reductions.
31897 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31898             Alan Hayward  <alan.hayward@arm.com>
31899             David Sherwood  <david.sherwood@arm.com>
31901         * tree.h (build_index_vector): Declare.
31902         * tree.c (build_index_vector): New function.
31903         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
31904         of units as polynomial, forcibly converting it to a constant if
31905         vectorizable_reduction has already enforced the condition.
31906         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
31907         to create a {1,2,3,...} vector.
31908         (vectorizable_reduction): Treat the number of units as polynomial.
31909         Choose vectype_in based on the largest scalar element size rather
31910         than the smallest number of units.  Enforce the restrictions
31911         relied on above.
31913 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31914             Alan Hayward  <alan.hayward@arm.com>
31915             David Sherwood  <david.sherwood@arm.com>
31917         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
31918         number of units as polynomial.
31920 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31921             Alan Hayward  <alan.hayward@arm.com>
31922             David Sherwood  <david.sherwood@arm.com>
31924         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
31925         * target.def (autovectorize_vector_sizes): Return the vector sizes
31926         by pointer, using vector_sizes rather than a bitmask.
31927         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
31928         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
31929         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
31930         Likewise.
31931         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
31932         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
31933         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
31934         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
31935         * omp-general.c (omp_max_vf): Likewise.
31936         * omp-low.c (omp_clause_aligned_alignment): Likewise.
31937         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
31938         * tree-vect-loop.c (vect_analyze_loop): Likewise.
31939         * tree-vect-slp.c (vect_slp_bb): Likewise.
31940         * doc/tm.texi: Regenerate.
31941         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
31942         to a poly_uint64.
31943         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
31944         the vector size as a poly_uint64 rather than an unsigned int.
31945         (current_vector_size): Change from an unsigned int to a poly_uint64.
31946         (get_vectype_for_scalar_type): Update accordingly.
31947         * tree.h (build_truth_vector_type): Take the size and number of
31948         units as a poly_uint64 rather than an unsigned int.
31949         (build_vector_type): Add a temporary overload that takes
31950         the number of units as a poly_uint64 rather than an unsigned int.
31951         * tree.c (make_vector_type): Likewise.
31952         (build_truth_vector_type): Take the number of units as a poly_uint64
31953         rather than an unsigned int.
31955 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31956             Alan Hayward  <alan.hayward@arm.com>
31957             David Sherwood  <david.sherwood@arm.com>
31959         * target.def (get_mask_mode): Take the number of units and length
31960         as poly_uint64s rather than unsigned ints.
31961         * targhooks.h (default_get_mask_mode): Update accordingly.
31962         * targhooks.c (default_get_mask_mode): Likewise.
31963         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
31964         * doc/tm.texi: Regenerate.
31966 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31967             Alan Hayward  <alan.hayward@arm.com>
31968             David Sherwood  <david.sherwood@arm.com>
31970         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
31971         * omp-general.c (omp_max_vf): Likewise.
31972         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
31973         (expand_omp_simd): Handle polynomial safelen.
31974         * omp-low.c (omplow_simd_context): Add a default constructor.
31975         (omplow_simd_context::max_vf): Change from int to poly_uint64.
31976         (lower_rec_simd_input_clauses): Update accordingly.
31977         (lower_rec_input_clauses): Likewise.
31979 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31980             Alan Hayward  <alan.hayward@arm.com>
31981             David Sherwood  <david.sherwood@arm.com>
31983         * tree-vectorizer.h (vect_nunits_for_cost): New function.
31984         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
31985         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
31986         (vect_analyze_slp_cost): Likewise.
31987         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
31988         (vect_model_load_cost): Likewise.
31990 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31991             Alan Hayward  <alan.hayward@arm.com>
31992             David Sherwood  <david.sherwood@arm.com>
31994         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
31995         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
31996         from an unsigned int * to a poly_uint64_pod *.
31997         (calculate_unrolling_factor): New function.
31998         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
32000 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32001             Alan Hayward  <alan.hayward@arm.com>
32002             David Sherwood  <david.sherwood@arm.com>
32004         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
32005         from an unsigned int to a poly_uint64.
32006         (_loop_vec_info::slp_unrolling_factor): Likewise.
32007         (_loop_vec_info::vectorization_factor): Change from an int
32008         to a poly_uint64.
32009         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
32010         (vect_get_num_vectors): New function.
32011         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
32012         (vect_get_num_copies): Use vect_get_num_vectors.
32013         (vect_analyze_data_ref_dependences): Change max_vf from an int *
32014         to an unsigned int *.
32015         (vect_analyze_data_refs): Change min_vf from an int * to a
32016         poly_uint64 *.
32017         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
32018         than an unsigned HOST_WIDE_INT.
32019         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
32020         (vect_analyze_data_ref_dependence): Change max_vf from an int *
32021         to an unsigned int *.
32022         (vect_analyze_data_ref_dependences): Likewise.
32023         (vect_compute_data_ref_alignment): Handle polynomial vf.
32024         (vect_enhance_data_refs_alignment): Likewise.
32025         (vect_prune_runtime_alias_test_list): Likewise.
32026         (vect_shift_permute_load_chain): Likewise.
32027         (vect_supportable_dr_alignment): Likewise.
32028         (dependence_distance_ge_vf): Take the vectorization factor as a
32029         poly_uint64 rather than an unsigned HOST_WIDE_INT.
32030         (vect_analyze_data_refs): Change min_vf from an int * to a
32031         poly_uint64 *.
32032         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
32033         vfm1 as a poly_uint64 rather than an int.  Make the same change
32034         for the returned bound_scalar.
32035         (vect_gen_vector_loop_niters): Handle polynomial vf.
32036         (vect_do_peeling): Likewise.  Update call to
32037         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
32038         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
32039         be constant.
32040         * tree-vect-loop.c (vect_determine_vectorization_factor)
32041         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
32042         (vect_get_known_peeling_cost): Likewise.
32043         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
32044         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
32045         (vect_transform_loop): Likewise.  Use the lowest possible VF when
32046         updating the upper bounds of the loop.
32047         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
32048         rather than an int.
32049         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
32050         polynomial unroll factors.
32051         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
32052         (vect_make_slp_decision): Likewise.
32053         (vect_supported_load_permutation_p): Likewise, and polynomial
32054         vf too.
32055         (vect_analyze_slp_cost): Handle polynomial vf.
32056         (vect_slp_analyze_node_operations): Likewise.
32057         (vect_slp_analyze_bb_1): Likewise.
32058         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
32059         than an unsigned HOST_WIDE_INT.
32060         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
32061         (vectorizable_load): Handle polynomial vf.
32062         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
32063         a poly_uint64.
32064         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
32066 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32067             Alan Hayward  <alan.hayward@arm.com>
32068             David Sherwood  <david.sherwood@arm.com>
32070         * match.pd: Handle bit operations involving three constants
32071         and try to fold one pair.
32073 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32075         * tree-vect-loop-manip.c: Include gimple-fold.h.
32076         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
32077         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
32078         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
32079         Add a path that uses a step of VF instead of 1, but disable it
32080         for now.
32081         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
32082         and niters_no_overflow parameters.  Update calls to
32083         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
32084         Create a new SSA name if the latter choses to use a ste other
32085         than zero, and return it via niters_vector_mult_vf_var.
32086         * tree-vect-loop.c (vect_transform_loop): Update calls to
32087         vect_do_peeling, vect_gen_vector_loop_niters and
32088         slpeel_make_loop_iterate_ntimes.
32089         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
32090         (vect_gen_vector_loop_niters): Update declarations after above changes.
32092 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
32094         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
32095         128-bit round to integer instructions.
32096         (ceil<mode>2): Likewise.
32097         (btrunc<mode>2): Likewise.
32098         (round<mode>2): Likewise.
32100 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
32102         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
32103         unaligned VSX load/store on P8/P9.
32104         (expand_block_clear): Allow the use of unaligned VSX
32105         load/store on P8/P9.
32107 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32109         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
32110         New function.
32111         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
32112         swap associated with both a load and a store.
32114 2018-01-02  Andrew Waterman  <andrew@sifive.com>
32116         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
32117         * config/riscv/riscv.md (clear_cache): Use it.
32119 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
32121         * web.c: Remove out-of-date comment.
32123 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32125         * expr.c (fixup_args_size_notes): Check that any existing
32126         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
32127         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
32128         (emit_single_push_insn): ...here.
32130 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32132         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
32133         (const_vector_encoded_nelts): New function.
32134         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
32135         (const_vector_int_elt, const_vector_elt): Declare.
32136         * emit-rtl.c (const_vector_int_elt_1): New function.
32137         (const_vector_elt): Likewise.
32138         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
32139         of CONST_VECTOR_ELT.
32141 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32143         * expr.c: Include rtx-vector-builder.h.
32144         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
32145         directly on the tree encoding.
32146         (const_vector_from_tree): Likewise.
32147         * optabs.c: Include rtx-vector-builder.h.
32148         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
32149         sequence of "u" values.
32150         * vec-perm-indices.c: Include rtx-vector-builder.h.
32151         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
32152         directly on the vec_perm_indices encoding.
32154 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32156         * doc/rtl.texi (const_vector): Describe new encoding scheme.
32157         * Makefile.in (OBJS): Add rtx-vector-builder.o.
32158         * rtx-vector-builder.h: New file.
32159         * rtx-vector-builder.c: Likewise.
32160         * rtl.h (rtx_def::u2): Add a const_vector field.
32161         (CONST_VECTOR_NPATTERNS): New macro.
32162         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
32163         (CONST_VECTOR_DUPLICATE_P): Likewise.
32164         (CONST_VECTOR_STEPPED_P): Likewise.
32165         (CONST_VECTOR_ENCODED_ELT): Likewise.
32166         (const_vec_duplicate_p): Check for a duplicated vector encoding.
32167         (unwrap_const_vec_duplicate): Likewise.
32168         (const_vec_series_p): Check for a non-duplicated vector encoding.
32169         Say that the function only returns true for integer vectors.
32170         * emit-rtl.c: Include rtx-vector-builder.h.
32171         (gen_const_vec_duplicate_1): Delete.
32172         (gen_const_vector): Call gen_const_vec_duplicate instead of
32173         gen_const_vec_duplicate_1.
32174         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
32175         (gen_const_vec_duplicate): Use rtx_vector_builder.
32176         (gen_const_vec_series): Likewise.
32177         (gen_rtx_CONST_VECTOR): Likewise.
32178         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
32179         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
32180         Build a new vector rather than modifying a CONST_VECTOR in-place.
32181         (handle_special_swappables): Update call accordingly.
32182         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
32183         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
32184         Build a new vector rather than modifying a CONST_VECTOR in-place.
32185         (handle_special_swappables): Update call accordingly.
32187 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32189         * simplify-rtx.c (simplify_const_binary_operation): Use
32190         CONST_VECTOR_ELT instead of XVECEXP.
32192 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32194         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
32195         the selector elements to be different from the data elements
32196         if the selector is a VECTOR_CST.
32197         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
32198         ssizetype for the selector.
32200 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32202         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
32203         before testing each element individually.
32204         * tree-vect-generic.c (lower_vec_perm): Likewise.
32206 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32208         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
32209         * selftest-run-tests.c (selftest::run_tests): Call it.
32210         * vector-builder.h (vector_builder::operator ==): New function.
32211         (vector_builder::operator !=): Likewise.
32212         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
32213         (vec_perm_indices::all_from_input_p): New function.
32214         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
32215         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
32216         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
32217         instead of reading the VECTOR_CST directly.  Detect whether both
32218         vector inputs are the same before constructing the vec_perm_indices,
32219         and update the number of inputs argument accordingly.  Use the
32220         utility functions added above.  Only construct sel2 if we need to.
32222 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32224         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
32225         the broadcast of the low byte.
32226         (expand_mult_highpart): Use an explicit encoding for the permutes.
32227         * optabs-query.c (can_mult_highpart_p): Likewise.
32228         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
32229         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32230         (vectorizable_bswap): Likewise.
32231         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
32232         explicit encoding for the power-of-2 permutes.
32233         (vect_permute_store_chain): Likewise.
32234         (vect_grouped_load_supported): Likewise.
32235         (vect_permute_load_chain): Likewise.
32237 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32239         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
32240         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
32241         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
32242         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
32243         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
32244         (vect_gen_perm_mask_any): Likewise.
32246 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32248         * int-vector-builder.h: New file.
32249         * vec-perm-indices.h: Include int-vector-builder.h.
32250         (vec_perm_indices): Redefine as an int_vector_builder.
32251         (auto_vec_perm_indices): Delete.
32252         (vec_perm_builder): Redefine as a stand-alone class.
32253         (vec_perm_indices::vec_perm_indices): New function.
32254         (vec_perm_indices::clamp): Likewise.
32255         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
32256         (vec_perm_indices::new_vector): New function.
32257         (vec_perm_indices::new_expanded_vector): Update for new
32258         vec_perm_indices class.
32259         (vec_perm_indices::rotate_inputs): New function.
32260         (vec_perm_indices::all_in_range_p): Operate directly on the
32261         encoded form, without computing elided elements.
32262         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
32263         encoding.  Update for new vec_perm_indices class.
32264         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
32265         the given vec_perm_builder.
32266         (expand_vec_perm_var): Update vec_perm_builder constructor.
32267         (expand_mult_highpart): Use vec_perm_builder instead of
32268         auto_vec_perm_indices.
32269         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
32270         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
32271         or double series encoding as appropriate.
32272         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
32273         vec_perm_indices instead of auto_vec_perm_indices.
32274         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
32275         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
32276         (vect_permute_store_chain): Likewise.
32277         (vect_grouped_load_supported): Likewise.
32278         (vect_permute_load_chain): Likewise.
32279         (vect_shift_permute_load_chain): Likewise.
32280         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
32281         (vect_transform_slp_perm_load): Likewise.
32282         (vect_schedule_slp_instance): Likewise.
32283         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32284         (vectorizable_mask_load_store): Likewise.
32285         (vectorizable_bswap): Likewise.
32286         (vectorizable_store): Likewise.
32287         (vectorizable_load): Likewise.
32288         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
32289         vec_perm_indices instead of auto_vec_perm_indices.  Use
32290         tree_to_vec_perm_builder to read the vector from a tree.
32291         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
32292         vec_perm_builder instead of a vec_perm_indices.
32293         (have_whole_vector_shift): Use vec_perm_builder and
32294         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
32295         truncation to calc_vec_perm_mask_for_shift.
32296         (vect_create_epilog_for_reduction): Likewise.
32297         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
32298         from auto_vec_perm_indices to vec_perm_indices.
32299         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
32300         instead of changing individual elements.
32301         (aarch64_vectorize_vec_perm_const): Use new_vector to install
32302         the vector in d.perm.
32303         * config/arm/arm.c (expand_vec_perm_d::perm): Change
32304         from auto_vec_perm_indices to vec_perm_indices.
32305         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
32306         instead of changing individual elements.
32307         (arm_vectorize_vec_perm_const): Use new_vector to install
32308         the vector in d.perm.
32309         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
32310         Update vec_perm_builder constructor.
32311         (rs6000_expand_interleave): Likewise.
32312         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
32313         (rs6000_expand_interleave): Likewise.
32315 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32317         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
32318         to qimode could truncate the indices.
32319         * optabs.c (expand_vec_perm_var): Likewise.
32321 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32323         * Makefile.in (OBJS): Add vec-perm-indices.o.
32324         * vec-perm-indices.h: New file.
32325         * vec-perm-indices.c: Likewise.
32326         * target.h (vec_perm_indices): Replace with a forward class
32327         declaration.
32328         (auto_vec_perm_indices): Move to vec-perm-indices.h.
32329         * optabs.h: Include vec-perm-indices.h.
32330         (expand_vec_perm): Delete.
32331         (selector_fits_mode_p, expand_vec_perm_var): Declare.
32332         (expand_vec_perm_const): Declare.
32333         * target.def (vec_perm_const_ok): Replace with...
32334         (vec_perm_const): ...this new hook.
32335         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
32336         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
32337         * doc/tm.texi: Regenerate.
32338         * optabs.def (vec_perm_const): Delete.
32339         * doc/md.texi (vec_perm_const): Likewise.
32340         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
32341         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
32342         expand_vec_perm for constant permutation vectors.  Assert that
32343         the mode of variable permutation vectors is the integer equivalent
32344         of the mode that is being permuted.
32345         * optabs-query.h (selector_fits_mode_p): Declare.
32346         * optabs-query.c: Include vec-perm-indices.h.
32347         (selector_fits_mode_p): New function.
32348         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
32349         is defined, instead of checking whether the vec_perm_const_optab
32350         exists.  Use targetm.vectorize.vec_perm_const instead of
32351         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
32352         fit in the vector mode before using a variable permute.
32353         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
32354         vec_perm_indices instead of an rtx.
32355         (expand_vec_perm): Replace with...
32356         (expand_vec_perm_const): ...this new function.  Take the selector
32357         as a vec_perm_indices rather than an rtx.  Also take the mode of
32358         the selector.  Update call to shift_amt_for_vec_perm_mask.
32359         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
32360         Use vec_perm_indices::new_expanded_vector to expand the original
32361         selector into bytes.  Check whether the indices fit in the vector
32362         mode before using a variable permute.
32363         (expand_vec_perm_var): Make global.
32364         (expand_mult_highpart): Use expand_vec_perm_const.
32365         * fold-const.c: Includes vec-perm-indices.h.
32366         * tree-ssa-forwprop.c: Likewise.
32367         * tree-vect-data-refs.c: Likewise.
32368         * tree-vect-generic.c: Likewise.
32369         * tree-vect-loop.c: Likewise.
32370         * tree-vect-slp.c: Likewise.
32371         * tree-vect-stmts.c: Likewise.
32372         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
32373         Delete.
32374         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
32375         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
32376         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
32377         (aarch64_vectorize_vec_perm_const): ...this new function.
32378         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32379         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32380         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
32381         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
32382         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32383         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32384         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
32385         into...
32386         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
32387         check for NEON modes.
32388         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
32389         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
32390         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
32391         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
32392         into...
32393         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
32394         the old VEC_PERM_CONST conditions.
32395         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
32396         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
32397         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
32398         (ia64_vectorize_vec_perm_const_ok): Merge into...
32399         (ia64_vectorize_vec_perm_const): ...this new function.
32400         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
32401         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
32402         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
32403         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
32404         * config/mips/mips.c (mips_expand_vec_perm_const)
32405         (mips_vectorize_vec_perm_const_ok): Merge into...
32406         (mips_vectorize_vec_perm_const): ...this new function.
32407         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
32408         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
32409         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
32410         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
32411         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
32412         (rs6000_expand_vec_perm_const): Delete.
32413         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
32414         Delete.
32415         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32416         (altivec_expand_vec_perm_const_le): Take each operand individually.
32417         Operate on constant selectors rather than rtxes.
32418         (altivec_expand_vec_perm_const): Likewise.  Update call to
32419         altivec_expand_vec_perm_const_le.
32420         (rs6000_expand_vec_perm_const): Delete.
32421         (rs6000_vectorize_vec_perm_const_ok): Delete.
32422         (rs6000_vectorize_vec_perm_const): New function.
32423         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
32424         an element count and rtx array.
32425         (rs6000_expand_extract_even): Update call accordingly.
32426         (rs6000_expand_interleave): Likewise.
32427         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
32428         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
32429         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
32430         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
32431         (rs6000_expand_vec_perm_const): Delete.
32432         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32433         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32434         (altivec_expand_vec_perm_const_le): Take each operand individually.
32435         Operate on constant selectors rather than rtxes.
32436         (altivec_expand_vec_perm_const): Likewise.  Update call to
32437         altivec_expand_vec_perm_const_le.
32438         (rs6000_expand_vec_perm_const): Delete.
32439         (rs6000_vectorize_vec_perm_const_ok): Delete.
32440         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
32441         reference to the SPE evmerge intructions.
32442         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
32443         an element count and rtx array.
32444         (rs6000_expand_extract_even): Update call accordingly.
32445         (rs6000_expand_interleave): Likewise.
32446         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
32447         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
32448         new function.
32449         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32451 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32453         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
32454         vector mode and that that mode matches the mode of the data
32455         being permuted.
32456         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
32457         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
32458         directly using expand_vec_perm_1 when forcing selectors into
32459         registers.
32460         (expand_vec_perm_var): New function, split out from expand_vec_perm.
32462 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32464         * optabs-query.h (can_vec_perm_p): Delete.
32465         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
32466         * optabs-query.c (can_vec_perm_p): Split into...
32467         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
32468         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
32469         particular selector is valid.
32470         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
32471         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
32472         (vect_grouped_load_supported): Likewise.
32473         (vect_shift_permute_load_chain): Likewise.
32474         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
32475         (vect_transform_slp_perm_load): Likewise.
32476         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32477         (vectorizable_bswap): Likewise.
32478         (vect_gen_perm_mask_checked): Likewise.
32479         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
32480         implementations of variable permutation vectors into account
32481         when deciding which selector to use.
32482         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
32483         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
32484         with a false third argument.
32485         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
32486         to test whether the constant selector is valid and can_vec_perm_var_p
32487         to test whether a variable selector is valid.
32489 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32491         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
32492         * optabs-query.c (can_vec_perm_p): Likewise.
32493         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
32494         instead of vec_perm_indices.
32495         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
32496         (vect_gen_perm_mask_checked): Likewise,
32497         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
32498         (vect_gen_perm_mask_checked): Likewise,
32500 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32502         * optabs-query.h (qimode_for_vec_perm): Declare.
32503         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
32504         (qimode_for_vec_perm): ...this new function.
32505         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
32507 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
32509         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
32510         does not have a conditional at the top.
32512 2018-01-02  Richard Biener  <rguenther@suse.de>
32514         * ipa-inline.c (big_speedup_p): Fix expression.
32516 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
32518         PR target/81616
32519         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
32520         for generic 4->6.
32522 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
32524         PR target/81616
32525         Generic tuning.
32526         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
32527         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
32528         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
32529         cond_taken_branch_cost 3->4.
32531 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
32533         PR tree-optimization/83581
32534         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
32535         TODO_cleanup_cfg if any changes have been made.
32537         PR middle-end/83608
32538         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
32539         convert_modes if target mode has the right side, but different mode
32540         class.
32542         PR middle-end/83609
32543         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
32544         last argument when extracting from CONCAT.  If either from_real or
32545         from_imag is NULL, use expansion through memory.  If result is not
32546         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
32547         the parts directly to inner mode, if even that fails, use expansion
32548         through memory.
32550         PR middle-end/83623
32551         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
32552         check for bswap in mode rather than HImode and use that in expand_unop
32553         too.
32555 Copyright (C) 2018 Free Software Foundation, Inc.
32557 Copying and distribution of this file, with or without modification,
32558 are permitted in any medium without royalty provided the copyright
32559 notice and this notice are preserved.